numerical modelling of the gas dynamics of a prototype free-piston engine · pdf...
TRANSCRIPT
Numerical Modelling of the Gas
Dynamics of a Prototype Free-Piston
Engine
Gregory Paul Gibbes
Submitted in fulfilment of the degree of
Doctor of Philosophy
University of Technology, Sydney
Australia
2011
keywords: free-piston, 1D gas dynamics, two-stroke, specific heat, chemical
equilibrium, charging, scavenging, tuning
ii
iii
CERTIFICATE OF AUTHORSHIP/ORIGINALITY
I certify that the work in this thesis has not previously been submitted for a degree
nor has it been submitted as part of requirements for a degree except as fully
acknowledged within the text.
I also certify that the thesis has been written by me. Any help that I have received in
my research work and the preparation of the thesis itself has been acknowledged. In
addition, I certify that all information sources and literature used are indicated in the
thesis.
……………………………..
iv
. . . .
v
ACKNOWLEDGEMENTS
I would first like to acknowledge the men who created Pempek Engine project.
Edward Wechner, who designed most of engine’s hardware, and Bert Van Der
Broek for his company’s provision of the Pempek scholarship and for his leadership.
Thanks also goes to Douglas Carter who designed the electrical systems, and who
was a constant, friendly, professional source of ideas and fruitful discussion. Thanks
to Slobadan Ilic, who was likewise a friendly and helpful member of the small
Pempek engine team, and who worked tirelessly to manufacture and assemble much
of the engine. I was glad for his encouragement, advice, and readiness to share
information.
My supevisor Guang Hong has been supportive over this long road, and has
provided much guidance, insight and advice which, I hope, has borne fruit in making
this thesis more useful to others. Thank you for your hard work to support me, even
when we didn’t always agree. I am in your debt.
Thanks also to Phyllis Agius for her ready helpfulness in all postgraduate student
matters, both to me and numerous other students.
Thanks to Matt Gaston, Peter Brady and John Reizes who were always generous
with their time, as I learned the art of CFD modelling early in this project.
Thanks go to the many people who I have learned from but never met except through
the pages of technical papers and text books. In particular, I would like to
acknowledge the late Gordon P. Blair, whose method of modelling 1D gas dynamics
I have adapted for this work. Thanks to Samuel J. Kirkpatrick who’s carefully
published experimental work was invaluable to me in validating my code. Thanks
also to John D. Anderson Jr. from whom I learned much about compressible flow;
and Markus Klein and Gary L. Borman in the field combustion modelling.
Thanks to Randy Lewis, a kindred spirit in the joys 1D gas dynamic modelling, who
provided valuable feedback for the material which appears in Chapter 4.
Thanks to my research friends here at UTS from all corners of the globe: Janitha
Wijesinghe, Reza Fathollahzadeh, Fabio Cumbe, Wade Smith, Ulrike Dackermann,
Debbie Marsh, Fook Choi, Minh Nguyen, Kifayah Amar, Dang Ho, Thanh Nguyen,
Jiping Niu, Wen Xing, Zhinous Zabihi, Xiaohang Pang, Ibrahim El-Saliby and
Sherub Phuntsho.
vi
. . . .
vii
ABSTRACT
Free-piston internal combustion engines found commercial success as air
compressors in the 1920’s and 1930’s, and afterward as gas turbine gasifiers for
stationary applications. Since that time they have failed to see commercial
application, however in the last decade or so there has been a resurgence of interest
in free-piston engines because of their ostensible simplicity and in the flexibility
afforded by an unconstrained piston.
This thesis reports the testing and modelling on a free-piston engine by Pempek
Systems Pty. Ltd. It is an opposed cylinder, electric machine, operating on a two
stroke cycle with direct fuel injection. Analysis of experimental cylinder pressure
shows that while compression ignition is suitably fast and reliable, the Pempek
engine suffers from (among other things) low charging efficiency. The aim of the
modelling work is to understand the reasons for this, and to investigate design
options for improvement.
A comprehensive, generally applicable 1D gas dynamics engine model has been
developed. The important features of this model are described in some detail. While
the model builds on existing methods, a number of unique contributions have been
made. A chemical equilibrium code was developed which is computationally
efficient and flexible. The 1D gas dynamics method is based on a method developed
at Queens University, Belfast (QUB) in the early 1990’s but has been thoroughly re-
worked in the way it handles friction, gas property changes and heat transfer. The
originally first order accurate method has been changed to second order, and a way
of preserving full mass conservation has been developed. An unsteady heat transfer
model is proposed. A comprehensive boundary solution is presented, which has
relevance to all 1D gas dynamics models. The gas dynamics model is validated
against extensive single shot data from QUB, and also against some experimental
engine-run data.
The 1D gas dynamics engine model is used to assess the viability of utilising exhaust
pipe tuning to drive the charging process of the Pempek engine. Simulation results
show that it is possible to charge the engine using exhaust gas dynamics alone.
viii
. . . .
ix
PREFACE
At a stand at the World Energy Congress 2004 in Sydney I met Bert Van Der Broek
and Edward Wechner and was introduced to the Pempek free-piston engine. Ed
explained to me the fascinating features of his new engine design and invited me to
visit the workshop to see some prototype testing. This was the start of my
association with Pempek Systems, which lead to a final year project exploring the
scavenging of the engine, then on to this PhD in a similar vein.
Pempek Systems had done what few research groups had yet achieved – they had
built and run a full scale electric free-piston engine, demonstrating unequivocally
that generator based piston motion control was accurate and robust. Their working
prototype was an excellent platform from which to launch a theoretical investigation
- which was aimed at providing tools to interpret results, and doing predictive
modelling to guide future directions of the project.
A little should be said about the contents of this thesis which follow from the
requirements of the project. There are two main topics addressed. The first is the
developing technology of free-piston engines. This is the subject of the first two
chapters. The second topic is that of engine modelling and makes up the middle part
of the thesis. Even though the modelling was developed for the Pempek project, it is
nonetheless broadly applicable to all IC engines and even to other fields. Thus, the
sections on modelling can be read profitably without concern for the preceding
chapters on free-piston engine technology. Likewise, readers with little interest in
physics and methods of modelling will be able to read the sections reporting free-
piston engine technology. The final section of the thesis takes the engine model and
looks at two possible design variations for the Pempek engine.
- Synopsis -
Chapter 1 - Free-piston Engines – overview of developments
Surveys the current state of the art in free-piston engine technology. The survey
shows that despite the relative immaturity of the field, promising solutions have been
found for the main difficulties, such as piston motion control.
Chapter 2 - Pempek free-piston engine – details and experimental results
Describes the Pempek free-piston engine project in some detail, highlighting the
successful piston motion control, and describing some of the difficulties that were
faced, in particular low combustion energy and high compressor power
Preface x
consumption. In order to explore the potential for lower compressor pressure, it was
deemed necessary to analyse the gas dynamics of inlet and exhaust systems. This is
the motivation for the modelling work which follows.
Chapter 3 - Thermodynamic and gas property models
Describes three key components of the engine model – namely the single zone
thermodynamic cylinder model, the gas property model and the chemical
equilibrium model.
Chapter 4 - Unsteady 1D gas dynamics model
Describes the unsteady gas dynamics model, which was based on an existing
method, but with several modifications. Concludes with some simple validation
cases.
Chapter 5 - Other sub models
Describes miscellaneous other parts of the engine model which were not covered in
the previous two chapters.
Chapter 6 - The engine model – integrating all sub models
Explains the integration of all the sub models into the overall engine model
Chapter 7 - Validations using experimental results
Validates the gas dynamics model against a suit of single shot experiments, and also
a superficial comparison to measured data from the Pempek engine.
Chapter 8 - Predictive modelling
Applies the gas dynamics model to the original Pempek engine but with a modified
low pressure compressor, and a tuned exhaust pipe. Simulation results show that
low compressor pressure operation is possible. Next, a radical design modification
is proposed, and the gas dynamics model is used to test the viability of un-boosted
charging. These two applications of the gas dynamics model demonstrate the
usefulness of the model, and the sort of design options that are available for free-
piston engines to take advantage of gas dynamics to improve and control charging.
Chapter 9 - Summary and conclusion
Summarises the specific findings for the Pempek project, summarises the model
scope and usefulness, and lists the unique contributions of the thesis. A list of
suggested further work is also included.
Preface xi
Appendices
A wide range of material with further details of free-piston engine projects and the
engine model.
Appendix I Review of recent free-piston engine projects
Appendix II Specific heats of a reacting mixture
Appendix III Further applications of the energy equation
Appendix IV Tables of Thermodynamic Properties
Appendix V Method for Calculating Chemical Equilibrium of Combustion Products
Appendix VI Derivation of fundamental one dimensional unsteady gas equation
Appendix VII Derivation of Boundary Flow Equations
Appendix VIII Model Data Structures
Appendix IX 2nd Order Interpolation – further details
Appendix X Re-meshing Criteria and Method
Appendix XI Single Shot Experiments Cross Reference
Appendix XII Derivation of normal shock equations
Appendix XIII Rayleigh and Fanno Flow
Appendix XIV Graphical user interface screen shots
Appendix XV Table of contents of data CD
xii
. . . .
xiii
TABLE OF CONTENTS
Abstract ................................................................................................................................. vii
Preface .................................................................................................................................... ix
Table of Contents................................................................................................................. xiii
List of Figures ..................................................................................................................... xvii
List of Tables ..................................................................................................................... xxvii
Nomenclature ..................................................................................................................... xxix
Acronyms ......................................................................................................................... xxxiii
Chapter 1 Free-piston Engines – overview of developments .......................................... 1
1.1 Introduction ............................................................................................................. 2
1.2 Summary of free-piston projects ............................................................................. 4
1.3 Piston Motion Control ............................................................................................. 6
1.4 Discussion on free-piston engines state of the art ................................................... 9
1.5 Free-piston Engine Modelling .............................................................................. 11
Chapter 2 Pempek free-piston engine – details and experimental results ................... 19
2.1 Overview of the project......................................................................................... 19
2.2 Details of the engine ............................................................................................. 21
2.3 Cylinder pressure analysis .................................................................................... 27
2.4 Compressor analysis ............................................................................................. 31
2.5 Summary ............................................................................................................... 33
Chapter 3 Thermodynamic and gas property models ................................................... 37
3.1 Thermodynamic control volume model ................................................................ 38
3.2 Gas mixture property model ................................................................................. 45
3.3 Reacting gas mixture model .................................................................................. 49
Chapter 4 Unsteady 1D gas dynamics model ................................................................. 55
4.1 Introduction ........................................................................................................... 55
4.2 Theoretical Basis ................................................................................................... 61
4.3 Wave Propagation ................................................................................................. 63
4.4 Flow Boundary solution ........................................................................................ 68
Table of Contents xiv
4.5 Mass and thermal energy transport ....................................................................... 79
4.6 Validation using analytical results ........................................................................ 83
4.7 Summary ............................................................................................................... 89
Chapter 5 Other sub models ............................................................................................ 91
5.1 Duct friction and heat transfer ............................................................................... 92
5.2 Combustion cylinder models ............................................................................... 105
5.3 Separated flow model .......................................................................................... 112
5.4 Flow area coefficient maps ................................................................................. 115
5.5 Multi body dynamics........................................................................................... 117
Chapter 6 The engine model – integrating all sub models .......................................... 123
6.1 Overview of model building blocks .................................................................... 124
6.2 Calculation sequence for a complete time-step evaluation ................................. 125
6.3 Programing details of the engine model .............................................................. 127
Chapter 7 Validations using experimental results ....................................................... 129
7.1 Description of the single shot tests ...................................................................... 130
7.2 Slide valve tests ................................................................................................... 134
7.3 P1 driven simulation ........................................................................................... 141
7.4 Straight pipe shots ............................................................................................... 142
7.5 Converging flow ................................................................................................. 147
7.6 Diverging flow .................................................................................................... 150
7.7 Modelling the Pempek engine ............................................................................. 161
Chapter 8 Predictive modelling ..................................................................................... 171
8.1 Optimising the existing layout ............................................................................ 172
8.2 Port admission layout .......................................................................................... 175
8.3 Gas dynamics driven scavenging ........................................................................ 178
8.4 Conclusion .......................................................................................................... 183
Chapter 9 Summary and conclusion ............................................................................. 185
9.1 Findings for the Pempek project ......................................................................... 186
9.2 The model, its scope and usefulness ................................................................... 187
9.3 Unique contributions to modelling art ................................................................ 188
9.4 Contribution to free-piston engine research ........................................................ 190
9.5 Further Work ....................................................................................................... 191
Table of Contents xv
Publications ......................................................................................................................... 193
References ........................................................................................................................... 195
Appendices .......................................................................................................................... 207
Appendix I Review of recent free-piston engine projects ................................................ 209
Appendix II Specific heats of a reacting mixture ............................................................. 235
Appendix III Further applications of the energy equation ............................................... 237
Appendix IV Tables of Thermodynamic Properties ........................................................ 241
Appendix V Method for Calculating Chemical Equilibrium of Combustion Products ... 247
Appendix VI Derivation of fundamental one dimensional unsteady gas equation .......... 255
Appendix VII Derivation of Boundary Flow Equations .................................................. 259
Appendix VIII Model Data Structures ............................................................................. 263
Appendix IX 2nd Order Interpolation – further details ..................................................... 269
Appendix X Re-meshing Criteria and Method ................................................................ 271
Appendix XI Single Shot Experiments Cross Reference ................................................. 273
Appendix XII Derivation of normal shock equations ...................................................... 275
Appendix XIII Rayleigh and Fanno Flow ........................................................................ 281
Appendix XIV Graphical user interface screen shots ...................................................... 285
Appendix XV Table of contents of data CD .................................................................... 291
xvi
. . . .
xvii
LIST OF FIGURES
Figure 1-1 Various free-piston engine layouts ............................................................ 2
Figure 1-2 Two control methodologies ....................................................................... 8
Figure 1-3 Scavenging modes analysed by Goldsborough [58] ................................ 15
Figure 1-4 Simulated 1D model compared to experiment for pressure in exhaust
pipe. Larmi et al [74] ....................................................................................... 17
Figure 2-1 Spark ignition prototype .......................................................................... 20
Figure 2-2 Cross section of engine module (shown vertically) ................................. 22
Figure 2-3 Typical exhaust valve actuator trajectory ................................................ 23
Figure 2-4 Real vs. target mover velocity ................................................................. 25
Figure 2-5 Example engine run indicated work (efficiency indicated by black
markers) ............................................................................................................ 26
Figure 2-6 Sample indicator plot – fired cycle .......................................................... 27
Figure 2-7 Sample indicator plot - motored cycle ..................................................... 27
Figure 2-8 Apparent heat release............................................................................... 28
Figure 2-9 Cylinder and compressor pressure during scavenging ............................ 29
Figure 2-10 Comparison of cylinder pressure with and without exhaust pipe .......... 30
Figure 2-11 Indicator plot of compressor .................................................................. 31
Figure 2-12 Proposed advanced passive inlet valve design [96] ............................... 32
Figure 3-1 Numerical solution of the energy equation .............................................. 40
Figure 3-2 Pressure history and error for various time steps ..................................... 42
Figure 3-3 Prescribed fuel burn rate and cylinder volume ........................................ 43
Figure 3-4 Pressure and temperature for combustion case ........................................ 43
Figure 3-5 Error for various time steps ..................................................................... 44
Figure 3-6 Specific heat Cp of common exhaust gas species .................................... 45
Figure 3-7 Typical pressure error incurred for setting =1.4 .................................... 45
Figure 3-8 Variation of with temperature and equivalence ratio for unburned and
burned mixture ................................................................................................. 47
List of Figures xviii
Figure 3-9 Equilibrium species mass fractions of a fuel-air mixture at various
temperatures and pressure ................................................................................ 52
Figure 3-10 Equilibrium species mass fractions of a fuel air mixture with varying
fuel to oxygen ratio ........................................................................................... 52
Figure 4-1 Evolving mass flow rate into an idealised duct ....................................... 55
Figure 4-2 A right travelling pressure wave .............................................................. 61
Figure 4-3 Oppositely moving pressure waves ......................................................... 62
Figure 4-4 Advancing pressure waves by one time step ........................................... 63
Figure 4-5 Second order interpolation of pressure waves ......................................... 64
Figure 4-6 Modifying pressure waves to account for heat transfer and mass
conservation ...................................................................................................... 65
Figure 4-7 Re-meshing a duct ................................................................................... 67
Figure 4-8 Detection of a travelling shock ................................................................ 67
Figure 4-9 Duct cell boundary nodes in space and time ........................................... 68
Figure 4-10 Variation of gas properties around a node in space and time ................ 68
Figure 4-11 Catalogue of all flow types considered .................................................. 70
Figure 4-12 A typical flow boundary showing all flow properties ........................... 72
Figure 4-13 Mass and thermal transport.................................................................... 79
Figure 4-14 Calculating boundary flow properties ................................................... 81
Figure 4-15 Temperature transport with different mixing coefficients ..................... 82
Figure 4-16 Standard shock tube results ................................................................... 84
Figure 4-17 Shock tube with mass conservation ....................................................... 85
Figure 4-18 Shock tube comparison between first and second order wave
interpolation ...................................................................................................... 86
Figure 4-19 Fanno flow ............................................................................................. 87
Figure 4-20 Rayleigh flow ........................................................................................ 87
Figure 4-21 Smearing of a triangular pulse traversing 100 mesh spaces .................. 88
Figure 5-1 Coefficient of friction for flow over a flat plate [90] ............................... 92
List of Figures xix
Figure 5-2 Fluid element experiencing friction ......................................................... 93
Figure 5-3 Fluid element experiencing heat transfer ................................................. 95
Figure 5-4 Heat transfer model compared to Dittus-Boelter for steady flow ............ 99
Figure 5-5 Heat transfer model turbulent kinetic energy for different turbulence
length scales ...................................................................................................... 99
Figure 5-6 Heat transfer model for a turbulence generating inlet ........................... 100
Figure 5-7 Heat transfer model compared to Dittus-Boelter for low speed flow .... 100
Figure 5-8 Heat transfer model for different cell spacing ....................................... 101
Figure 5-9 Heat transfer modelled for single shot using different turbulence length
scales............................................................................................................... 102
Figure 5-10 Heat transfer modelled for single shot using different timestep size ... 103
Figure 5-11 Sketch of piston-cylinder crevice ........................................................ 106
Figure 5-12 Blowby CFD model mesh ................................................................... 107
Figure 5-13 Blowby correlation .............................................................................. 108
Figure 5-14 Finding fuel injection enthalpy ............................................................ 109
Figure 5-15 Spark ignition combustion rate model ................................................. 110
Figure 5-16 Compression ignition combustion rate model ..................................... 111
Figure 5-17 Control volume for applying the momentum equation to a diffusing flow
........................................................................................................................ 112
Figure 5-18 Modified area ratio for tapered ducts................................................... 114
Figure 5-19 Example flow area coefficient map ..................................................... 116
Figure 5-20 Cutaway of FP3 showing moving parts ............................................... 117
Figure 5-21 Forces on the mover ............................................................................ 117
Figure 5-22 Forces on the exhaust valves ............................................................... 118
Figure 5-23 Aero force coefficients for normal flow through exhaust and inlet valves
........................................................................................................................ 119
Figure 5-24 Forces on the passive inlet valves ........................................................ 120
Figure 5-25 Typical collision trajectory .................................................................. 122
List of Figures xx
Figure 6-1 Integration of sub models to make an engine model ............................. 123
Figure 7-1 Straight pipe........................................................................................... 130
Figure 7-2 Straight pipe with density discontinuity ................................................ 131
Figure 7-3 Sudden contraction ................................................................................ 131
Figure 7-4 Convergent taper ................................................................................... 131
Figure 7-5 Sudden enlargement .............................................................................. 132
Figure 7-6 Divergent taper ...................................................................................... 132
Figure 7-7 Short megaphone ................................................................................... 132
Figure 7-8 Long megaphone ................................................................................... 132
Figure 7-9 Flow area coefficients used for slide valve ........................................... 135
Figure 7-10 Slide valve Prel =1.5 bar, Trel=293K ..................................................... 136
Figure 7-11 Slide valve Prel =1.5 bar, Trel=293K, air in cylinder, CO2 in pipe ....... 136
Figure 7-12 Slide valve Prel =1.5 bar, Trel=293K, CO2 in cylinder, air in pipe........ 136
Figure 7-13 Slide valve Prel =2.4 bar, Trel=293K ..................................................... 137
Figure 7-14 Slide valve Prel =2.4 bar, Trel=293K, air in cylinder, CO2 in pipe........ 137
Figure 7-15 Slide valve Prel =2.4bar, Trel=293K, CO2 in cylinder, air in pipe......... 137
Figure 7-16 Slide valve Prel =2 bar, Trel=623K ........................................................ 138
Figure 7-17 Slide valve Prel =5 bar, Trel=623K ........................................................ 138
Figure 7-18 Slide valve Prel =0.5 bar, Trel=293K ..................................................... 138
Figure 7-19 Slide valve Prel =0.8 bar, Trel=293K ..................................................... 139
Figure 7-20 Slide valve Prel =0.5 bar, Trel=293K short pipe shot ............................ 139
Figure 7-21 Slide valve Prel =0.8 bar, Trel=293K short pipe shot ............................ 139
Figure 7-22 Slide valve Prel =2.4 bar, Trel=293K short pipe shot ............................ 140
Figure 7-23 Straight pipe P2, Prel =0.5 bar, Trel=293K ............................................ 143
Figure 7-24 Straight pipe P2, Prel =0.8 bar, Trel=293K ............................................ 143
Figure 7-25 Straight pipe P2, Prel =1.5 bar, Trel=293K ............................................ 143
Figure 7-26 Straight pipe P2, Prel =2.4 bar, Trel=293K ............................................ 144
List of Figures xxi
Figure 7-27 Density discontinuity P3, CCC, Prel =2.4 bar, Trel=293K, closed end . 144
Figure 7-28 Density discontinuity P1, AAC, Prel =1.5 bar, Trel=293K .................... 144
Figure 7-29 Density discontinuity P3, AAC, Prel =1.5 bar, Trel=293K .................... 144
Figure 7-30 Density discontinuity P1, AAC, Prel =2.4 bar, Trel=293K .................... 145
Figure 7-31 Density discontinuity P3, AAC, Prel =2.4 bar, Trel=293K .................... 145
Figure 7-32 Density discontinuity P1, CCA, Prel =1.5 bar, Trel=293K .................... 145
Figure 7-33 Density discontinuity P3, CCA, Prel =1.5 bar, Trel=293K .................... 145
Figure 7-34 Density discontinuity P1, CCA, Prel =2.4 bar, Trel=293K .................... 146
Figure 7-35 Density discontinuity P3, CCA, Prel =2.4 bar, Trel=293K .................... 146
Figure 7-36 Flow area coefficients used for sudden area change ........................... 147
Figure 7-37 Sudden contraction 53mm P1, Prel =2.4bar, Trel=293K ....................... 148
Figure 7-38 Convergent taper 53mm P1, Prel =2.4bar, Trel=293K........................... 148
Figure 7-39 Sudden contraction 53mm, P3, Prel =2.4bar, Trel=293K ...................... 148
Figure 7-40 Convergent taper 53mm, P3, Prel =2.4bar, Trel=293K .......................... 148
Figure 7-41 Sudden contraction 80.2mm, P1, Prel =2.4bar, Trel=293K ................... 149
Figure 7-42 Convergent taper 80.2mm, P1, Prel =2.4bar, Trel=293K ....................... 149
Figure 7-43 Sudden contraction 80.2mm, P3, Prel =2.4bar, Trel=293K ................... 149
Figure 7-44 Convergent taper 80.2mm, P3, Prel =2.4bar, Trel=293K ....................... 149
Figure 7-45 Sudden enlargement 53mm, P1, Prel =1.5bar, Trel=293K .................... 152
Figure 7-46 Divergent taper 53mm, P1, Prel =1.5bar, Trel=293K ............................ 152
Figure 7-47 Sudden enlargement/ Divergent taper 53mm, P3, Prel =1.5bar,
Trel=293K ........................................................................................................ 152
Figure 7-48 Sudden enlargement 53mm, P1, Prel =2.4bar, Trel=293K .................... 153
Figure 7-49 Divergent taper 53mm, P1, Prel =2.4bar, Trel=293K ............................ 153
Figure 7-50 Sudden enlargement/ Divergent taper 53mm, P3, Prel =2.4bar,
Trel=293K ........................................................................................................ 153
Figure 7-51 Sudden enlargement 80.2mm, P1, Prel =1.5bar, Trel=293K ................. 154
Figure 7-52 Divergent taper 80.2mm, P1, Prel =1.5bar, Trel=293K ......................... 154
List of Figures xxii
Figure 7-53 Sudden enlargement/ Divergent taper 80.2mm, P3, Prel =1.5bar,
Trel=293K ........................................................................................................ 154
Figure 7-54 Sudden enlargement 80.2mm, P1, Prel =2.4bar, Trel=293K ................. 155
Figure 7-55 Divergent taper 80.2mm, P1, Prel =2.4bar, Trel=293K ......................... 155
Figure 7-56 Sudden enlargement/ Divergent taper 80.2mm, P3, Prel =2.4bar,
Trel=293K ........................................................................................................ 155
Figure 7-57 Divergent taper 80.2mm, P2, Prel =2.4bar, Trel=293K ......................... 156
Figure 7-58 Divergent taper 105.6mm, P1, Prel =2.4bar, Trel=293K ....................... 156
Figure 7-59 Divergent taper 105.6mm, P2, Prel =2.4bar, Trel=293K ....................... 156
Figure 7-60 Divergent taper 105.6mm, P3, Prel =2.4bar, Trel=293K ....................... 157
Figure 7-61 Short Megaphone P1, Prel =2.0bar, Trel=293K ..................................... 157
Figure 7-62 Long Megaphone P1, Prel =2.0bar, Trel=293K ..................................... 157
Figure 7-63 Sudden contraction 53mm, P1, Prel =0.5bar, Trel=293K ...................... 158
Figure 7-64 Convergent taper 53mm, P1, Prel =0.5bar, Trel=293K.......................... 158
Figure 7-65 Sudden contraction / Convergent taper 53mm, P3, Prel =0.5bar,
Trel=293K ........................................................................................................ 158
Figure 7-66 Sudden contraction 53mm, P1, Prel =0.8bar, Trel=293K ...................... 159
Figure 7-67 Convergent taper 53mm, P1, Prel =0.8bar, Trel=293K.......................... 159
Figure 7-68 Sudden contraction / Convergent taper 53mm, P3, Prel =0.8bar,
Trel=293K ........................................................................................................ 159
Figure 7-69 Short Megaphone P1, Prel =2.0bar, Trel=293K, various simulation
timesteps ......................................................................................................... 160
Figure 7-70 Long Megaphone P1, Prel =2.0bar, Trel=293K, various simulation
timesteps ......................................................................................................... 160
Figure 7-71 Pempek engine inlet side ducting half section 3D view ...................... 161
Figure 7-72 Pempek engine1D model layout .......................................................... 162
Figure 7-73 Motored engine comparison with experiment ..................................... 163
Figure 7-74 Motored engine, indicator diagram ..................................................... 164
Figure 7-75 Motored engine, indicator diagram ..................................................... 164
List of Figures xxiii
Figure 7-76 Fired engine comparison with experiment .......................................... 165
Figure 7-77 Fired engine, indicator diagram ........................................................... 165
Figure 7-78 Fired engine, indicator diagram ........................................................... 166
Figure 7-79 Valve trajectories and mass flows ....................................................... 168
Figure 7-80 Cylinder mass and inlet mass .............................................................. 168
Figure 7-81 Cylinder species mass fractions ........................................................... 169
Figure 7-82 Cylinder temperature and specific heat ratio ....................................... 169
Figure 8-1 Tuned exhaust pipe layout for existing engine ...................................... 172
Figure 8-2 Results for optimised layout .................................................................. 173
Figure 8-3 Valve trajectories and mass flows (full power) ..................................... 173
Figure 8-4 Cylinder mass and inlet mass (full power) ............................................ 174
Figure 8-5 Cylinder species mass fractions (full power)......................................... 174
Figure 8-6 Modification of Pempek engine for port admission .............................. 175
Figure 8-7 Contactless pistons with air bearings ..................................................... 177
Figure 8-8 Port admission engine model layout ...................................................... 178
Figure 8-9 General results for port admission layout .............................................. 179
Figure 8-10 Port openings and mass flows (full power) ......................................... 180
Figure 8-11 Cylinder pressure during scavenging (full power) .............................. 180
Figure 8-12 Modelled inlet and exhaust plenum pressure (full power) .................. 180
Figure 8-13 Cylinder mass and inlet mass (full power) .......................................... 181
Figure 8-14 Cycle temperatures .............................................................................. 181
Figure 8-15 Cycle pressures .................................................................................... 182
Figure 8-16 Cycle pressures during scavenging ...................................................... 182
Figure I-1 Junkers four-stage free-piston air compressor [8] .................................. 210
Figure I-2 Partial cut-away diagram of the SIGMA Type GS-34 Free-Piston Gasifier
[8] ................................................................................................................... 211
Figure I-3 INNAS “Chiron” hydraulic free-piston engine [7] ................................ 212
List of Figures xxiv
Figure I-4 Tampere University of Technology hydraulic free-piston engine prototype
“Emma2” [113] .............................................................................................. 213
Figure I-5 Third Generation prototype engine [114] ............................................... 214
Figure I-6 Cutaway view of the Toyohashi university of Technology hydraulic free-
piston engine [63] ........................................................................................... 215
Figure I-7 EPA six-cylinder four-stroke FPE [32] .................................................. 216
Figure I-8 West Virginia University, second generation linear engine prototype (2-
stroke) [115] ................................................................................................... 218
Figure I-9 West Virginia University, four stroke concept [97] ............................... 219
Figure I-10 First Sandia Free-piston Linear Alternator concept [116].................... 220
Figure I-11 Sandia opposed piston layout [119] ..................................................... 221
Figure I-12 Sandia bounce chamber detail showing compressed air injection valves
on the left and vent ports on the right [119] ................................................... 222
Figure I-13 Sandia Opposed Piston Free-piston Engine [120] ................................ 223
Figure I-14 Timeline of Sandia free-piston engine project [119]............................ 223
Figure I-15 FPEC prototype electric generator – opposed cylinder, two-stroke diesel
with pnumatic exhaust valves [78] ................................................................. 224
Figure I-16 Korea Institute of Energy opposed cylinder engine [125] ................... 225
Figure I-17 Four-stroke free-piston engine concept by Nanjing University of
Science and Technology [128] ....................................................................... 227
Figure I-18 German Aerospace Center free-piston prototype on test bench [60] ... 227
Figure I-19 Linear alternator built at WVU [34] ..................................................... 228
Figure I-20 Sandia linear alternator design [117] ................................................... 229
Figure I-21 Magnaquench linear alternator stator [117] ......................................... 229
Figure I-22 University of Sheffield FPEC linear permanent magnet generator [122]
........................................................................................................................ 230
Figure I-23 Royal Institute of Technology Stockholm, Sweden transverse flux
permanent magnet generator [40] ................................................................... 230
Figure I-24 German Aerospace Center linear generator on the test stand [60] ....... 231
List of Figures xxv
Figure I-25 Design concept for a miniature HCCI free-piston engine [112] .......... 232
Figure I-26 Photo of Kvaerner test cylinder unit .................................................... 233
Figure I-27 Liquid piston air compressor prototype (high inertance model)
Vanderbilt University [73].............................................................................. 234
Figure III-1 Variation of R with temperature for products of combustion in chemical
equilibrium ..................................................................................................... 240
Figure V-1 Heat release of n-octane at various equivalence ratios and temperatures
........................................................................................................................ 253
Figure VI-1 A fluid element influenced by a pressure wave ................................... 255
Figure VII-1 Schematic of a duct boundary ............................................................ 259
Figure IX-1 Interpolating discontinuous pressure waves ........................................ 269
Figure IX-2 Handling the duct ends ........................................................................ 270
Figure IX-3 Single cell ducts .................................................................................. 270
Figure XII-1 Schematic of a normal shock ............................................................. 275
Figure XII-2 Travelling shock ................................................................................. 280
Figure XIV-1 Main screen ...................................................................................... 286
Figure XIV-2 Edit Ducts screen .............................................................................. 286
Figure XIV-3 Edit Volumes screen ......................................................................... 287
Figure XIV-4 Edit Bodies screen ............................................................................ 287
Figure XIV-5 Edit Area Coefficients screen ........................................................... 288
Figure XIV-6 Edit Functions screen ....................................................................... 288
Figure XIV-7 Create Animated plot screen ............................................................ 289
Figure XIV-8 Example Animation screen grab (shock tube problem) ................... 289
Figure XIV-9 History plot screen ............................................................................ 290
xxvi
. . . .
xxvii
LIST OF TABLES
Table 3-1 Thermodynamic cylinder model calculation sequence ............................. 41
Table 3-2 List of species considered in equilibrium calculation ............................... 50
Table 3-3 Equilibrium equations ............................................................................... 51
Table 4-1 Shock tube setup ....................................................................................... 83
Table IV-1 Enthalpy of some combustion products ................................................ 241
Table IV-2 Specific heat of some combustion products ......................................... 242
Table IV-3 Properties of some fuels (various sources [30, 62, 107]) ...................... 243
Table IV-4 Equilibrium equations ........................................................................... 244
Table IV-5 Equilibrium constants for selected reactions ........................................ 245
Table XIV-1 Cross reference figure numbers for single shot data .......................... 273
xxviii
. . . .
xxix
NOMENCLATURE
Symbols
A area (m2)
a speed of sound (m/s)
a0 isentropic reference speed of sound (m/s)
a acceleration (m/s2)
moles of atomic element i.e. (mol)
circumference or wetted perimeter of a duct (m)
Coefficient of friction (Fanning friction factor) (-)
Coefficient of heat transfer (W/m2/K)
specific heat at constant pressure and constant volume (J/kg/K)
specific heat at constant pressure and constant volume (J/mol/K)
frozen specific heats (see Appendix II) (J/kg/K)
c wave velocity (m/s)
hydraulic diameter (m)
F force (N)
Enthalpy, enthalpy flow rate (J, J/s)
specific enthalpy, enthalpy of formation (J, J/kg, J/mol)
thermal conductivity (W/m/K)
Length
Large eddy length
M momentum (kg.m/s)
M molar mass (g/mol)
mass, mass flow rate (kg, kg/s)
Total mixture moles (mol)
Species moles (mol)
Nomenclature xxx
P, P0 pressure, reference pressure (absolute pressure, Pa)
heat transfer rate (J/s, J/kg/s)
gas constant (J/kg/K)
universal gas constant (8.314472 J/mol/K)
Reynolds Number, turbulent Reynolds number (-)
T temperature (K)
isentropic reference temperature (K)
t time (s)
u, u0 fluid velocity, quiescent fluid velocity (m/s)
turbulence intensity, RMS of fluctuating velocity (m/s)
internal energy, specific internal energy (J, J/kg, J/mol)
v velocity (m/s)
V, v volume, specific volume (m3, m3/kg)
work rate (J/s), specific work rate (J/kg/s),
21
0PPX pressure amplitude ratio (-)
species mole fraction (-)
ratio of specific heats (-)
ratio of frozen specific heats (-)
viscosity (N.s/m2)
Courant number (-)
density (kg/m3)
etc species molar concentration (kmol/m3)
Nomenclature xxxi
Subscripts
L leftward
R rightward
i iteration, incident
f flow
prev previous
r reflected
s species
xxxii
. . . .
xxxiii
ACRONYMS
BDC Bottom dead centre
BTDC Before top dead centre
CFD Computational Fluid Dynamics
FPEC Free-piston Energy Converter, project name for a European free-
piston engine consortium
HCCI Homogenous Charge Compression Ignition
IMEP Indicated mean effective pressure
QUB Queens University Belfast
TDC Top dead centre
WRT With respect to
xxxiv
. . . .
1
Chapter 1 Free-piston Engines – overview of
developments
Free-piston combustion engines found commercial success as air compressors in the
1920’s and 1930’s, and then as gas turbine gasifiers for stationary applications.
However, since that time they have failed to see commercial application. There is
sporadic literature in the intervening years on several renewed attempts, and in the
last decade or so there has been a resurgence of interest in the idea.
This chapter gives an outline for the state of free-piston engine development at this
time. Basic engine layout options are summarised, recent published engine projects
are listed, piston motion control is discussed, and the general state of the art is
critiqued. Next, several of the more notable modelling efforts are described, with
emphasis on modelling related to the modelling conducted here. An in-depth
description of the Pempek engine project and experimental results follows in Chapter
2.
Chapter 1 Free-piston Engines – overview of developments 2
1.1 Introduction
Free-piston engine layouts can be usefully categorised by piston-cylinder
arrangement. The three basic possibilities are shown in Figure 1-1.
Single ended layouts have only one combustion cylinder, and must have some way
of returning the piston for the next compression stroke. Single piston hydraulic
machines can have lighter movers which result in higher speed and higher power
output. For hydraulic cases, the single ended layout can allow pulse pause control.
Electric machines usually use a so-called gas spring to return the piston. The mass
of air in the bounce chamber can be varied to help control piston motion.
Figure 1-1 Various free-piston engine layouts
In the opposed cylinder arrangement, there are two identical combustion chambers
and a single moving assembly, and a centrally located pump or generator. This
arrangement has the advantage of not needing a bounce mechanism to return the
piston at BDC, since the opposite cylinder achieves this function. The engine is
symmetrical, as is the motion of the mover. This means that the trajectory at BDC
must be the same as the trajectory at TDC.
Opposed piston machines are essentially two single ended machines placed so that
their combustion pistons occupy either end of the same combustion cylinder. Thus
there are two generators/pumps and two bounce mechanisms. These machines enjoy
Opposed cylinder
Single ended
Opposed piston
generator or pump cylinder
piston
cylinder
Bounce mechanism
Chapter 1 Free-piston Engines – overview of developments 3
all of the advantages and suffer the disadvantages of single ended machines with the
additional advantage of highly effective uni-flow port scavenging being possible,
intrinsic dynamic balance, and increased effective stroke to bore ratio. At the same
time, there is the added constraint on the control system of keeping the pistons
carefully synchronised.
Other possible layouts include a four stroke engine made up of two opposed cylinder
machines with a common mover such as proposed by [97], or the six cylinder layout
described in [32]. The eight cylinder two stroke engine proposed by [33] is made of
four opposed cylinder modules arranged so that the inertia of the four synchronised
movers cancel.
Most free-piston engine concepts operate on a two stroke cycle. The two stroke
cycle relies on rapid transfer of cylinder gases around the bottom of the piston stroke
during which time both inlet and exhaust ports/valves may be open.
If the fuel is pre-mixed with air outside the cylinder, then fuel short circuiting must
be avoided by careful management of the scavenging process, such as using a
reduced scavenging ratio. This in turn means that it is impossible to obtain a
homogenous cylinder charge and several researchers have found spark ignition to be
somewhat unreliable due to high residual exhaust gas fraction. In practice
compression ignition has often been used to give reliable combustion, even when
scavenging is incomplete.
Direct fuel injection allows higher scavenging ratios to be employed, since only pure
air may short circuit the cylinder. High speed, high pressure common rail fuel
injectors which allow multiple injections during the compression and combustion
process open opportunities to create customised fuel mixture patterns.
The speed at which scavenging must be carried out generally necessitates some form
of pressurisation on the inlet air, and this is usually carried out by means of a
separate compressor, or an integrated crank-case-style compression volume.
A number of researchers have concluded that due to difficulties of the two stroke
cycle it was worth pursuing a mechanically more-challenging four stroke concept.
Eg [32, 45, 97, 128].
Chapter 1 Free-piston Engines – overview of developments 4
1.2 Summary of free-piston projects
The following brief overview of free piston engine projects is not exhaustive, and
focuses on recently reported projects.
1.2.1 Early free-piston engines
Free-piston engines are not new [8]. Prolific inventor Raúl Pateras Pescara
developed a number of designs for free-piston air compressors beginning in 1922. At
around the same time the German company Junkers began developing similar air
compressors, which were used to provide compressed air for torpedo launch tubes on
German submarines. The next major phase in free-piston technology was introduced
in 1944 with the Pescara designed 600 kW SIGMA gasifiers. In the intervening
years several attempts have been made at producing smaller gasifiers, such as the
General Motors GMR 4-4 Hyprex but without success [8].
1.2.2 Hydraulic free-piston engines
Following the gasifier type free-piston engine, hydraulic engines have been
developed to various degrees, however none have yet (to the author’s knowledge)
been produced commercially. Hydraulic free-piston engines utilise hydraulic oil to
extract piston work. The earlier success of hydraulic engines compared to electric
free-piston engines can be attributed to the ubiquitous nature of high performance
hydraulic componentry compared to the less advanced field of high performance
linear electric machines, and the relatively recent arrival of high coercivity
permanent magnets and appropriate solid state electric power technology. In
addition, some researchers suggest there is less difficulty in adequately controlling
hydraulic free-piston machines. The following overview lists hydraulic free-piston
engines published since about the year 2000 onwards.
The Chiron engine by Dutch company INNAS is a single ended, direct injection
compression ignition machine, and was an advanced prototype in 2000 [7]. There
has been little further word since that time. Tampere and Helsinki Universities of
Technology designed and built several opposed cylinder compression ignition
prototypes [114]. Toyohashi University of Technology have had a long running
project based on an opposed piston uni-flow port scavenged machine [63]. The U.S.
EPA and NVFEL engaged in an intensive free-piston engine development program
as part of their hydraulic vehicle program. Two stroke and four stroke variants were
Chapter 1 Free-piston Engines – overview of developments 5
developed, both based on an opposed cylinder, compression ignition layout. Exhaust
was via hydraulic or cam actuated poppet valves [32]. Beijing Institute of
Technology have a single ended hydraulic machine under development [126].
1.2.3 Electric free-piston engines
West Virginia University have had a long running free-piston engine project with
several prototypes reported [64, 91, 115]. All have been opposed cylinder spark
ignition machines. Sandia National Laboratories have had an ongoing free-piston
engine project which grew out of research into hydrogen and high compression ratio
engines. Initially an opposed cylinder machine was designed [116], but more
recently an opposed piston machine has been constructed with a view to testing a
range of different fuels [120]. A consortium of companies and universities obtained
European funding to develop the Free-piston Energy Converter (FPEC). It is an
opposed cylinder diesel machine with pneumatic exhaust valves [78], and is
probably the most advanced electric free-piston engine to date. The University of
Newcastle upon the Tyne began a free-piston engine project in 1999. More recently,
numerous papers have been published detailing various modelling efforts for a
proposed two stroke single ended machine [80]. Australian company Pempek
Systems developed and tested a compact two-stroke opposed cylinder machine [33].
Details of this project are given in Chapter 2. Loughborough University partnered
with Sheffield University and Lotus Engineering Ltd. in 2005 under a UK
government EPSRC grant to develop a four-stroke free-piston engine [45]. Other
electric free-piston engine projects are reported from the Korea Institute of Energy
[125], Malaysian Ministry of Science, technology and Environment [17], Shanghai
Jiao Tong University [75], Nanjing University of Science and Technology [128], and
the German Aerospace Center [60].
***
The above catalogue of recent and current free-piston engine projects demonstrates
the renewed activity in this field in the past decade. Greater detail on these and other
projects is provided in Appendix I, including an overview of notable linear generator
projects.
Chapter 1 Free-piston Engines – overview of developments 6
1.3 Piston Motion Control
Unlike a crank driven piston, a free-piston travels under the combined influence of
gas pressure and any load forces. The question arises as to whether this combination
of forces results in inherently stable or unstable reciprocations. The question of
engine control is central to recent work by Mikalsen and Roskilly [80-88] which
focuses on electric free-piston engines. Their modelling work discovered that in the
absence of suitable closed loop control input, the compression ratio from stroke to
stoke is unstable.
Intriguingly however, some practical free-piston engines have been reported to
operate stably with no closed loop control. The hydraulic free-piston engine from
Tampere University of Technology was capable of steady operation at a fixed
hydraulic load and fixed fuelling level [113], and was more stable in practice than
modelling suggested [74]. All of the prototypes reported from West Virginia
University [64, 91, 115] have been run at constant fuel level and variable load. Even
more surprisingly, the load used in these cases was a friction brake which is not a
viscous-type load such as a passive alternator would be. This author believes that
the demonstrated stability of the WVU machines is mainly due to the damping
influence of combustion chamber leakage (crevice volumes and blowby). The
stability in the Tamper machine was conditional on steady load and is likely to be the
combined result of combustion chamber losses and hydraulic losses at over-stroke
conditions.
Nevertheless, the majority of recent researchers have reported machines with
feedback control. The free-piston gasifier from Kvaerner ASA [66, 67] uses a
combination of fuel level, valve timing and bounce chamber pressure to achieve
piston control. Tikkanen and Vilenius [114] report a control system designed for
operating a refined version of the hydraulic machine that was originally run
uncontrolled. Control is affected by varying fuel mass. Information about changes
in the hydraulic load is used by the controller to pre-empt the resulting change to
piston motion and allow rapid response in changes to fuelling level. The controller
was tested on a Matlab/Simulink model of the engine and showed good ability to
control compression ratio, provided that the rate of change in load level was kept
within certain limits. No experimental results were reported. The hydraulic free-
piston engine proposed by the Beijing Institute of Technology [126] will also try to
achieve piston control by varying fuel levels on a stroke by stroke basis.
Chapter 1 Free-piston Engines – overview of developments 7
In contrast to using fuel level as a control input, the hydraulic machines developed
by the US EPA [32] achieved piston motion control by varying the hydraulic power
extraction on a stroke by stroke basis in response to changes in fuel energy.
Controllable check valves in the hydraulic circuit were utilised to provide this
control and allow arbitrary hydraulic outlet pressure. The Toyohashi University of
Technology machine [63] uses pulse-pause modulation to control power output.
Each cycle is independent, thus disturbances in one cycle are not carried over to the
next.
In the field of electric free-piston engines, the experimental free-piston engine rig
being developed at Sandia [120] uses compressed air automatically injected into the
bounce chambers to supplement combustion energy on a stroke by stroke basis. In
its present form with a fixed alternator load, this control system would have
excessive energy requirements at low load, however in the future a controllable
alternator load may alleviate this problem. The most advanced electric free-piston
engine project presently reported is the Free-piston Energy Converter (FPEC). This
machine uses a controlled electric machine allowing electromagnetic force to be
used as a control input [78]. The Pempek free-piston engine and the engine being
developed at the German Aerospace Center [60] also uses generator force as the
control input. The ambitious four stroke concept proposed at the Nanjing University
of Science and Technology [128] relies heavily on a bi-directional electric machine
to control and drive the motion of the spring mounted piston.
Mikalsen and Roskilly’s extensive numerical investigations on free-piston engine
control assume the load imposed on the mover is not a control input, though they do
acknowledge the possibility of implementing some kind of alternator control through
the use of power electronics (indeed controllability of the machine is assumed for the
purposes of starting the engine [80]). The controllability (or lack thereof) of the
mover load turns out to be a crucial differentiating factor in all of the control
strategies reported. They produce two entirely distinct control philosophies.
Figure 1-2 illustrates the two different control methods. In Figure 1-2(a) the load on
the engine is set by the user and the engine control system controls fuel injection
duration and timing to regulate piston motion and attempt to maintain steady
compression ratio. Single ended or opposed piston machines may also control gas
pressure in a bounce chamber to regulate piston motion. The problem with this
method is that there is a time delay of about one stroke before control action can be
taken. It also depends on low combustion variability. Mikalsen and Roskilly
Chapter 1 Free-piston Engines – overview of developments 8
consider piston control to be the main un-resolved challenge for the free-piston
engine concept and propose a predictive motion control method to reduce this time
delay and improve the transient response of the engine [88].
The second control philosophy is illustrated in Figure 1-2(b). The fuel quantity is set
by the user and the engine control system controls the load imposed on the mover to
regulate piston motion and attempt to maintain steady compression ratio. This
method results in demonstrably superior piston motion control since irregularities in
motion can be compensated almost immediately by a change in load, and the control
is not reliant on predictable combustion. However it requires the generator or pump
to be a controlled machine, not a passive machine.
Figure 1-2 Two control methodologies
Free-piston
engine
User set load
Fuel etc Piston motion
Control
Free-piston
engine
User set fuel
load Piston motion
Control
(a)
(b)
Chapter 1 Free-piston Engines – overview of developments 9
1.4 Discussion on free-piston engines state of the art
Despite the variety of free-piston concepts, all of the practitioners of this field are
attracted to the potential benefits of free-piston engines which can be summarised as:
Intrinsic variable compression ratio
Ostensible mechanical simplicity
The intrinsic variable compression ratio is due to the nature of the free-piston which
is un-constrained. In comparison to the kinematic nature of a crank driven piston,
whose path is determined by the geometry of the mechanism, the free-piston follows
a kinetic trajectory which can be adjusted by both changes in cylinder pressure and
load. Compression ratio is a critical parameter in the combustion process, and the
opportunity to choose an appropriate compression ratio for differing fuels and load
points is very attractive. Robust control of compression ratio has already been
demonstrated by several machines (eg Pempek, U.S. EPA [32]) but control of
machines without load control still appears problematic. A commonly stated aim of
free-piston prototype development is to implement Homogenous Charge
Compression Ignition (HCCI) which offers high efficiency and low emissions. This
requires fairly precise control of cylinder conditions to control the combustion
timing, and variable compression ratio can be used to accomplish this (providing that
accurate compression ratio control is achievable). Not only can combustion be
easily adapted across the load range using variable compression ratio, but variations
in fuel quality and even fuel type can be accommodated too. The fact that free-
pistons spend significantly less time near “top dead centre” than conventional
cranked pistons means that higher compression ratios are possible before the onset of
auto ignition, and heat transfer and NOx production are likely to be lowered [87]
(though combustion must be fast enough to avoid “time loss”). Finally, several
researchers are advocating very high compression ratio combustion (CR>30:1) as a
path to efficiency improvement [108, 109, 121].
The mechanical simplicity of free-piston engines has been noted by many
researchers, though it must be said that this does not necessarily translate to simple
construction. Although there are only one or two main moving parts, many engines
also have various valving requirements. For electric free-piston engines, a low
weight, efficient linear electric machine remains a technological challenge. Support,
lubrication and cooling of the piston assembly has received little if any attention in
the literature, but as more projects get beyond the concept stage and begin running
Chapter 1 Free-piston Engines – overview of developments 10
prototypes for long periods, lubrication and wear will become an important issue for
the mechanical design. Lubrication of piston rings is problematic for ported engines
(which most reported engines are), and may lead to increased burning of lubricating
oil in the cylinder. On the other hand, free-pistons have relatively low side loads to
support, and as several researchers have pointed out, higher compression ratios can
be more easily achieved with free-pistons compared to a cranked piston which
requires large bearings and stiff blocks and shafts to constrain the piston at high
compression ratios. Another possibility that has only been superficially explored in
relation to micro engines is un-lubricated ringless pistons that rely on very close
piston-cylinder tolerances and alignment for gas sealing.
Chapter 1 Free-piston Engines – overview of developments 11
1.5 Free-piston Engine Modelling
There is little that makes modelling of free-piston engines distinct from modelling
cranked piston engines. The only significant difference is due to piston motion.
Prescribed piston motion based on crank geometry is the common method for
cranked engines, and this is sufficiently accurate for most purposes. However free-
piston motion is a complex function of cylinder pressure(s) and instantaneous mover
load, so a truly predictive model needs to solve the piston dynamics coupled with
cylinder pressure and load.
Most modelling tools developed for cranked engines can be successfully applied to
free-piston engines, though certain empirical equations may need to be modified.
Nevertheless, the following section will summarise some notable efforts at
modelling free-piston engines in recent times.
1.5.1 Piston Dynamics Model
The motion of the rigid piston or mover assembly can be described using the
momentum equation as
where is any force that is exerted on the mover. Integrating WRT time yields
change in velocity, and integrating again yields change in position. Since the various
forces that act on the mover are complex and depend on non-linear influences such
as such combustion, valve/port opening and load behaviour, it is inadvisable to seek
an analytical solution. Rather, the problem can be conveniently solved on a
computer on a time-stepping basis where each force is evaluated instant by instant.
Since the initial value of velocity influences the solution, the solution will in general
vary somewhat for each cycle. If the dynamic behaviour of the engine model - run at
steady settings - tends toward stability, then the mover trajectory will settle to a
steady state after a number of engine cycles. Shoukry [105] reports his engine model
converged to a steady state after approximately 50 cycles, however my own
experience with the Pempek free-piston engine model was for around 3-6 cycles.
After the cylinder pressure, the most influential force on the mover is the load force.
Goldsborough and Van Blarigan [59] model the linear alternator as a force
proportional to velocity where the proportionality constant is constant throughout a
cycle but can be changed for different operating conditions. The WVU free-piston
Chapter 1 Free-piston Engines – overview of developments 12
engine model introduced by Petreanu [97] modelled the alternator load as a
sinusoidal function of position with a variety of shapes. The function was designed
to extract a set amount of work per stroke, which could be set by the user.
Friction has normally been found to be a small contributor to the mover motion,
owing to the low side force on the pistons compared to conventional engines.
Goldsborough and Van Blarigan [59] model friction as a combination of static and
viscous components with coefficients correlated to previous experiments. Shoukry
[105] describes a detailed friction model based on analysis of piston rings.
The monolithic liquid piston design recently developed at Vanderbilt University
[129] was modelled as a mass-spring-damper system, while the subsequent increased
inertance design used a flow based equation [123].
The largest contributor to piston motion is usually the cylinder pressure, especially at
the extremity of stroke when the pressure is very high. The determination of this
pressure is crucial for a viable engine model.
1.5.2 Simplified Cylinder Models
Single zone thermodynamic models form the basis of most free-piston engine piston
dynamics models. These treat the cylinder as a homogenous control volume which
is analysed using some variation of the energy equation. This is a powerful and
simple analysis which can yield accurate results. The accuracy of the model depends
on accurate specification of boundary flows, heat transfer, combustion rate and gas
mixture properties. A selection of single zone models is described below.
Goldsborough and Van Blarigan [59] modelled a free-piston engine using a
homogenous single zone. Combustion was modelled using a detailed chemical
kinetic mechanism in the hope of achieving better combustion model accuracy than a
reduced mechanism. The energy equation was written in terms of the species mass
fractions, internal energies and their rates of change. Combustion chamber leakage
was neglected. Heat transfer was by the Woschni correlation. Gas exchange was
modelled using equations for orifice flow, and a zonal formulation was employed to
model scavenging.
Petreanu [97] describes a cylinder model which operates in four differing phases.
Compression (until combustion begins) and expansion (after combustion ends) are
modelled as polytropic processes with a pre-determined polytropic index. The
combustion phase is modelled using the energy equation and assuming constant ratio
Chapter 1 Free-piston Engines – overview of developments 13
of specific heats. Heat transfer (presumably only applied during combustion) is
modelled using the Woschni correlation. Chemical heat release is modelled as part
of the heat transfer term and is prescribed with a Wibe function for mass fraction
burned. The fourth phase, the gas exchange period beginning with exhaust port
opening, was set to intake pressure, and the gas was instantaneously replaced with
cool, inlet air or air-fuel mixture (assuming perfect scavenging).
It is not immediately clear why the author used a polytropic exponent model instead
of the energy equation for the whole cycle, though perhaps it was deemed easier to
adjust to match experimental data. Regardless of the rational, this method restricts
the ability to use the model predictively, since the polytropic exponent represents an
empiricism requiring determination a priori. Also, the model ignores the gas
exchange process which, though bearing little immediate influence on piston motion,
effects important cylinder charge properties - namely residual exhaust fraction,
initial temperature and initial pressure. Nevertheless, this idealised gas exchange
model is sufficient to find an approximation of piston dynamics.
Larmi et al [74] report a cylinder model for a hydraulic engine using the energy
equation and accounted for inlet and exhaust flows using the nozzle flow equation,
heat transfer using the Woschni correlation and combustion using a triple-Wiebe
function. No mention was made of any gas blowby (leakage) model. The model
was capable of predicting the combustion cylinder and scavenging case pressures
with good accuracy, however air intake pressure had to be fudged higher, and
exhaust port pressure lower, since gas dynamic effects were apparently influencing
these processes. The model assumed fixed heat release and fixed hydraulic load, and
was rather less stable in operation than the real engine – with slight changes to the
load causing a cascading increase or decrease in piston velocity and compression
ratio. The author concludes that a more detailed hydraulic load model is needed.
Aichlmayr [12] developed a more sophisticated chemical kinetics single zone
cylinder model for modelling HCCI combustion in a micro free-piston cylinder. The
model used the energy equation and included the effect of blowby (which was
significant in this case). A good match with experimental data was achieved.
Mikalsen and Roskilly implement a simple cylinder model using the energy
equation, assuming constant specific heats and no blowby [80].
Single zone models have certain limitations because they cannot represent charge
inhomogeneities which affect combustion and emissions etc. Multi-zone models are
Chapter 1 Free-piston Engines – overview of developments 14
not common in free-piston modelling, perhaps because not many free-piston engines
use spark ignition; instead they typically employ either HCCI or direct injection
compression ignition. For greater combustion model fidelity in these cases a CFD
cylinder model is probably more appropriate.
1.5.3 Detailed cylinder models
The details of actual in-cylinder flows can be modelled using CFD, allowing
improved estimation of combustion chemistry and rate, emissions, heat transfer and
scavenging. However they are much more computationally intensive compared to
single and multi-zone models; and mesh setup can be complex and labour intensive.
Gas exchange modelling is also limited by the accuracy with which intake and
exhaust duct flows are modelled.
In an effort to find an optimum layout for scavenging, Goldsborough and Van
Blarigan [58] used the CFD code KIVA to analyse several configurations – loop,
hybrid-loop and uni-flow scavenging (Figure 1-3). Reasons for wanting good
scavenging were to prevent early ignition (due to added heat of charge mixture),
minimise fuel short circuiting and to ensure fast, reliable combustion.
Loop and hybrid-loop scavenging were both found to be unsatisfactory due to poor
scavenging performance, so a uni-flow layout was adopted. Optimum scavenging
was found with slow inlet velocity, zero port inclination and a swirl angle of 15
degrees. In the model, piston dynamics were prescribed based on previous 0D
modelling.
Kleemann et al [70] conducted 2D axisymetric and 3D simulations on the free-piston
engine of the FPEC consortium. They used a CFD code derived from KIVA-II but
considerably extended for modelling fuel injection and HCCI combustion. The
simulations were carried out iteratively between a 0D engine model, a 1D gas
dynamics model and the detailed CFD models.
Chapter 1 Free-piston Engines – overview of developments 15
Figure 1-3 Scavenging modes analysed by Goldsborough [58]
In a more detailed CFD analysis on the same project, Fredriksson et al [51] used
KIVA-3V to simulate the in-cylinder scavenging flows and combustion of a direct
injection diesel free-piston engine. A diesel oil surrogate model was implemented
into the KIVA code. Detailed chemical oxidation mechanisms of the two surrogate
fuels were modelled with 70 species in 306 reactions. Piston trajectory was
prescribed on the basis of previous 0D modelling. One significant finding was that
initial cylinder flow conditions (scavenging) had a large impact on subsequent
combustion, necessitating at least two or three complete cycles to gain realistic initial
conditions. Change of injection schedule had a significant influence on combustion
performance.
Mikalsen and Roskilly used the open source CFD code OpenFOAM to model the
combustion chamber of spark ignition free-piston engine compared to a conventional
cranked engine to analyse the difference in NO and CO emissions [81]. Initial
mixture after exhaust port closing seems to have been assumed homogenous and a
certain swirl velocity was prescribed. The piston trajectories were prescribed based
on previous analysis. Combustion was modelled using a flame area approach. The
concentration of NO and CO was modelled according to chemical kinetics equations
and for all other species, chemical equilibrium was assumed. A similar analysis was
also undertaken for DI diesel combustion [83]. These investigations included heat
transfer but no blowby. The model was subsequently extended to allow free-piston
dynamics to be directly coupled with cylinder pressure, allowing the effect of
combustion timing on compression ratio to be analysed [84]. The effect of an
Chapter 1 Free-piston Engines – overview of developments 16
insulated combustion chamber on emissions and efficiency is also investigated using
this model [87].
Mau et al [76] used commercial CFD code AVL_FIRE to model the cylinder and
“crank-case” of an electric free-piston engine under development to analyse the
differences in scavenging performance for a range of piston strokes and frequencies,
port positions and supercharging pressures. Piston motion was prescribed based on a
0D cylinder model and linear alternator electromagnetic model.
CFD models of the engine cylinder are dependent on the initial or boundary
conditions – in particular initial mixture distribution, temperature, velocity and initial
pressure. These can be predicted if the ducting leading up to the cylinder is
modelled, but the computational cost of extending the flow domain of CFD models
to the entire inlet and exhaust ductwork is prohibitive. A useful solution is to utilise
a simplified 1D gas dynamics code to provide time varying boundary conditions near
the cylinder inlets and outlets. 0D cylinder models are equally dependant on
boundary conditions, and here too a 1D gas dynamics model can be used to predict
cylinder inflows and outflows with greater realism than fixed pressure boundaries.
1.5.4 Gas dynamics
Reports of gas dynamics modelling in free-piston engines are few. This is perhaps a
reflection of the early state of development of many of these projects, where
predictions of combustion, scavenging, piston motion and valve timing have
assumed primary importance. However gas dynamic effects (especially in the two
stroke engine) tend to be highly significant, modifying flow rates through ports and
valves and affecting the final cylinder pressure. The breathing ability of a two stroke
engine can be dramatically improved and supercharging requirements reduced by
careful design of engine ducting. Goldsborough and Van Blarigan recognised the
importance of a gas dynamic analysis and planned to do it following a CFD
scavenging analysis [58]. A few 1D engine modelling efforts are outlined below.
Larmi et al [74] identified gas dynamic effects in experimental results of hydraulic
free-piston engine, compared to a 0D model of the same engine. A 1D gas dynamics
model was built using the commercial package GT-Power. Since this package is
designed for use with cranked engines, a number of work-arounds had to be
employed to enable it to model a free-piston engine. The piston motion had to be
specified in a lookup table as a function of crank angle degrees (CAD). This meant
Chapter 1 Free-piston Engines – overview of developments 17
that the piston motion was not dynamically calculated but constrained to a pre-set
trajectory, based on previous 0D simulations. The port openings likewise had to be
modelled as throttle valves opening as functions of CAD since the standard port
opening calculation was not suited to the asymmetric piston trajectory of a free-
piston engine. The simulated results of their work approximately followed
measured exhaust pipe pressure (Figure 1-4), thought it was unable to capture the
high frequency pulsations present in the measured data.
Figure 1-4 Simulated 1D model compared to experiment for pressure in exhaust
pipe. Larmi et al [74]
Fredriksson and Denbratt [52] also modelled the FPEC using commercial 1D gas
dynamics code BOOST. The cylinder model in the code was replaced by detailed
chemistry calculations from the SENKIN code.
Kleemann et al [70] mention a 1D gas dynamics model in their iterative 0D-1D-3D
modelling process, but do not give any details.
1.5.5 Other modelling approaches
Tóth-Nagy used a simulation model in conjunction with a combined genetic
algorithm-artificial neural network predictor model to explore the parameter space of
a free-piston engine for optimum combinations [115].
18
. . . .
19
Chapter 2 Pempek free-piston engine – details and
experimental results
2.1 Overview of the project
The Pempek free-piston engine is a symmetrical opposed cylinder, two stroke,
electric machine. It features an integrated compressor, overlapping generator
structure, and passive intake valves located in the head of the piston. [33, 96] The
design is targeted at compactness as it was envisaged for installation in hybrid
electric vehicles.
Pempek Systems Pty. Ltd. was engaged in developing this concept from about 2001
to 2008. During this period, the design of the original concept underwent minor
modification, and several versions were built and tested.
The first complete prototype used a single passive inlet valve in each piston, a single
electro pneumatic exhaust valve in each head, direct cylinder injection of gasoline
and spark ignition. The generator was a buried permanent magnet type, with a total
of eight poles on the mover and eight coils in the stator, each of which was actively
controlled by IGBT power transistors. A photo of the assembled spark ignition
prototype is shown in Figure 2-1. Engine testing commenced in early 2005.
Inconsistent combustion suggested scavenging issues, and the single large piston
mounted inlet valve was suspected of creating a large pocket of residual exhaust gas
in the cylinder. A four inlet valve piston was built to replace the single valve piston,
and the internal air space in the piston was reduce to ensure high scavenge air
pressure. At around the same time, a new cylinder head was introduced with a
central high pressure diesel injector and four exhaust valves.
This configuration proved to be much more successful at producing consistent
combustion. However, despite strong combustion and good indicated work, the
generator output remained stubbornly low. This was attributed to various loss
mechanisms in the generator, and a completely new generator was designed in an
effort to radically improve efficiency. The new generator employed a laminated
steel stator (the original used a powdered iron matrix), three times the number of
coils and a more traditional surface mounted magnet topology with back iron. The
same generator diameter was retained, however a decrease in the required magnet
Chapter 2 Pempek free-piston engine – details and experimental results 20
thickness allowed the combustion cylinder diameter to be slightly increased, and the
total mover mass was also decreased.
Figure 2-1 Spark ignition prototype
Unfortunately several problems plagued the new generator such as regular coil
failure, and lack of surface integrity of the inner stator surface, which caused rapid
wear of the mover bearings and imprecise positioning of the piston in the
combustion cylinder. Due to these problems, the final prototype was never
completely tested. Therefore, the following summary of experimental results will
draw on the successful engine runs from the original diesel configuration.
Two other un-tested modifications were proposed by the designer for this engine: A
low pressure passive inlet valve mechanism (see Figure 2-12), and an advanced low
energy electromagnetic exhaust valve actuator.
spark plug
generator stator body
gasoline injector
air intakes
exhaust valve actuator
cylinder head
generator coil terminals
Chapter 2 Pempek free-piston engine – details and experimental results 21
2.2 Details of the engine
A cross section view of the Pempek engine is shown in Figure 2-2. Note that the
engine operates in a horizontal position, and is shown vertically here because of
space constraints. This diagram shows the new generator, however the
specifications listed below belong to the original diesel configuration.
2 cylinders per module
dimensions ~150x150x700 mm per module
Cylinder diameter 66mm
maximum stroke 112mm
typical stroke 104mm
typical geometric compression ratio 26:1
mover mass 5.97kg
operating frequency ~31Hz (1920rpm equivalent)
indicated power output ~12kW per module
instrumentation
o cylinder pressure sensor1
o compressor pressure sensor
o exhaust valve actuator cavities pressure sensor
o mover position2
o control signals log
The diesel version also had the following specifications
Common rail injector from Mercedes E320 CDI
common rail fuel pressure 200-1400 bar adjustable
1 The cylinder pressure sensor was an un cooled optic fibre type by Optrand mounted in the cylinder head. Published accuracy for combustion applications is 2% (FSO). The sensor package includes the signal conditioning unit, which outputs an analogue voltage or current proportional to measured pressure. Measured pressure data showed significant long term drift from one cycle to the next (cumulatively several bar), though this stabilised after the first few engine cycles. As a result, it was necessary to arbitrarily peg pressure each cycle to some assumed value. Roth et al [103] Roth, K., Sobiesiak, A., Robertson, L. A., and Yates, S., 2002, In-Cylinder Pressure Measurements With Optical Fiber and Piezoelectric Pressure Transducers, SAE, Paper 2002-01-0745 reported a detailed study of the Optrand sensor-signal conditioner package in combustion applications. They reported that it has a rather complex response to thermal shock, over-estimating the peak pressure, then under-estimating the remainder of the expansion stroke pressure. Peak under-estimation during scavenging could be as much as 0.5bar according to their results. 2 The calculation of move position was implemented by Pempek as part of the engine control system. The estimated accuracy was +-0.5mm.
Chapter 2 Pempek free-piston engine – details and experimental results 22
Figure 2-2 Cross section of engine module (shown vertically)
exhaust valves
mover assembly
stator coil
cylinder sleeve
passive inlet valves
combustion cylinder
exhaust valve actuator
diesel injector
compressor intake check valve
compressor volume
diesel injector
exhaust valve armature
exhaust valve actuator coil
compressor cross-over duct
return springs
piston
cylinder head
cylinder head and exhaust port
air inlet
cylinder pressure sensor location
compressor pressure sensor location
Chapter 2 Pempek free-piston engine – details and experimental results 23
The unbalanced vibration of a single two cylinder module can be cancelled out if
four modules are run together synchronously with opposite piston phasing.
The most unusual feature of the Pempek engine is the inlet air path. As shown in
Figure 2-2, air is drawn into the compressor through a non-return valve. The mover
assembly acts as a compressor and pressurises the inlet air. The compressed air is
delivered to a cavity within the opposite piston, before being released into the
cylinder via passive inlet valves in the head of the piston. This novel design feature
allows the engine to avoid using intake ports in the cylinder wall which can cause
accelerated piston ring wear and elevated oil consumption.
A significant space saving innovation is the overlapping of the large generator
magnet holder with the combustion cylinders.
A lot of effort went into the development of fast-acting electro-pneumatic exhaust
valve actuators. These were capable of opening and closing in 7-10ms. There were
some reliability problems however, relating to leakage of the gas spring cavities.
Furthermore, electric power consumption was quite high. (But new designs were
proposed to overcome these problems).
A typical opening trajectory of the quad exhaust valve armature is shown in Figure
2-3. The motion of the armature is inferred1 from the measured gas spring pressure.
Figure 2-3 Typical exhaust valve actuator trajectory
The exact motion of the exhaust valves themselves is slightly different to the above
trajectory, because they are spring mounted to the armature. (see Figure 5-22 below)
1 The closing gas spring chamber pressure, along with the known closed valve position volume of the chamber, allowed the valve armature position to be calculated by assuming polytropic compression/expansion and a polytropic exponent n=1.4.
-1
0
1
2
3
4
5
6
0 0.002 0.004 0.006 0.008 0.01 0.012
Lift
(mm
)
time (s)
Chapter 2 Pempek free-piston engine – details and experimental results 24
In contrast to the exhaust valves, the intake valves (mounted in the piston head) open
passively when cylinder pressure is lower than compressor pressure. No direct
measurements were made of the opening trajectory of the intake valves, due to their
inaccessible location. However subsequent modelling shows their motion is highly
dependent on the combination of gas pressure forces in the cylinder and compressor,
and on the acceleration of the piston.
The exhaust valve opening and closing signals, fuel injection and spark timing (in
the case of the spark ignition configuration) were managed by a customised control
system, and were based on mover position. The position of the mover was not
directly measured but could be estimated based on generator coil voltage data.
The generator was an actively controlled machine, enabling it to regulate the motion
of the mover and control compression ratio. A typical velocity vs. position plot of
the mover is shown in Figure 2-4. The generator control system continually
monitored the mover velocity and position and adjusted the load force according to
whether the mover velocity was above or below the target velocity as
(2-1)
where is the instantaneous mover velocity and is a proportional control constant,
typically chosen to be about . The target velocity is determined as a
function of mover position as
(2-2)
where is the peak target velocity, is the target excursion from the
centred position, and the exponents and can be chosen to modify the shape of the
target velocity trajectory. The target trajectory is also shown in Figure 2-4. Note
that the centre position for the mover is nominated as the zero position.
Chapter 2 Pempek free-piston engine – details and experimental results 25
Figure 2-4 Real vs. target mover velocity
This control method has proven to be very effective at controlling compression ratio.
The same control algorithm can be used for starting and motoring the engine.
A typical engine run is shown in Figure 2-5. The engine was first started and
motored for eight strokes (four cycles). Fuel injection began on the ninth stroke, and
the test ran for a further 13 cycles. Most of the engine tests were kept deliberately
short to protect the generator which (as mentioned above) was known to be
dissipating a large amount of energy.
-10
-8
-6
-4
-2
0
2
4
6
8
10
-60 -40 -20 0 20 40 60
velo
city
(m/s
)
mover position (mm)
target
measured
Chapter 2 Pempek free-piston engine – details and experimental results 26
Figure 2-5 Example engine run indicated work (efficiency indicated by black
markers)
The indicated work for each cycle is shown in this example. While the engine is
motoring, heat and mass loss in the cylinder absorb around 50J per cycle. Once
firing, the energy of combustion produces positive work of around 150J per cycle.
Unfortunately, in this case, the left cylinder has a malfunctioning exhaust valve
which is effecting scavenging in this cylinder. On strokes 28 and 30, the valve
actually fails to open at all, and no combustion can then follow since there is no fresh
air in the cylinder. The important thing to note here is that despite wide fluctuations
in combustion energy in the left cylinder, the generator control system is easily able
to maintain correct compression. More detailed analysis of the engine performance
is presented below.
stroke
Indi
cate
d W
ork
(J)
Indi
cate
d ef
ficie
ncy
assu
min
g 14
.5m
g of
fuel
(%)
Chapter 2 Pempek free-piston engine – details and experimental results 27
2.3 Cylinder pressure analysis
Figure 2-6 and Figure 2-7 show sample indicator plots from the right hand cylinder.
(See footnote on page 21 for details of the pressure sensor used, and Figure 2-2 for
sensor location) In this instance, injection of diesel fuel was at 300 bar (relatively
low) and beginning when the mover was 48mm from centre (about 1.6ms BTDC).
The fired cycle indicated work is 167 joules, or IMEP 4.7 bar, equivalent to about 10
kW for both cylinders. Based on the estimated fuel quantity of 14.5mg and a fuel
LHV of 43MJ/kg, the indicated efficiency is 27%.
The motored indicated work in this case was -57joules, or IMEP 1.6 bar, equivalent
to about -3.5 kW for both cylinders. Evidence here points to high leakage from the
combustion chamber, causing the high energy consumption for motoring, and
contributing to the poor indicated efficiency for the fired cycle.
Figure 2-6 Sample indicator plot – fired cycle
Figure 2-7 Sample indicator plot - motored cycle
0
10
20
30
40
50
60
70
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
Pres
sure
(bar
)
mover position (mm)
167 J
0
10
20
30
40
50
60
70
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
Pres
sure
(bar
)
mover position (mm)
-57 J
Chapter 2 Pempek free-piston engine – details and experimental results 28
Figure 2-8 shows the net apparent heat release rate (HRR) from the cycle shown in
Figure 2-6. It was calculated very simply as
(2-3)
where the ratio of specific heats was assumed .
Figure 2-8 Apparent heat release
The above results can only be considered approximate, since a more rigorous
analysis would require modelling of the cylinder gas composition and temperature to
determined varying specific heats. Furthermore, actual combustion heat release
requires the effect of heat transfer and mass loss to be included. See Appendix III
for a further discussion on thermodynamic cylinder modelling.
The large dip in HRR just before the mixture ignites is mainly due to mass leakage
(though heat transfer and fuel evaporation also contribute). The main combustion
event lasts for approximately 3ms in total (relatively slow since the injection
pressure is only 300bar). Oddly, the results show continued heat release throughout
the expansion stroke, suggesting perhaps ongoing mixing controlled combustion (or
sensor drift may instead be to blame).
The estimated injected fuel quantity in this case was 14.5mg. Assuming complete
combustion, the actual heat release should be around 610J. Given that the apparent
heat release is only about 360J, it is clear that there are major losses - the most likely
-60
-40
-20
0
20
40
60
-100
0
100
200
300
400
500
-0.003 -0.001 0.001 0.003 0.005 0.007 0.009 0.011
posi
tion
(mm
)
HRR,
ene
rgy
(kJ/
s, J)
time (s)
mover position TDC
net apparent heat release (J)
heat release rate (kJ/s)
Chapter 2 Pempek free-piston engine – details and experimental results 29
being incomplete combustion (due to insufficient oxygen and poor
mixing/evaporation), and mass leakage (blowby).
Figure 2-9 shows the cylinder pressure during the scavenging period. The exhaust
valves are seen to open rather early when the expansion stroke is about 70%
complete. This is followed soon after by the opening of the passive inlet valves, as
cylinder pressure falls below compressor pressure. Since the pressure sensor for the
compressor is located some distance from the intake valves (see Figure 2-2), there is
a short delay before the measured compressor pressure begins to fall. As fresh air
floods into the cylinder through the opening intake valves (from 13ms-16ms) there is
a momentary rise in cylinder pressure, followed by a dip. Due to the type of
pressure sensors used, the absolute pressure here is unknown and must be guessed.
Figure 2-9 Cylinder and compressor pressure during scavenging
Things to notice here
compressor pressure over three times atmospheric pressure
fluctuating cylinder pressure during scavenging (up to 0.4 bar)
unknown inlet valve trajectory (though it can be inferred to some extent)
The high compressor pressure was deliberately designed to ensure that the inlet
valves would open correctly and that scavenging would be strongly driven. A closer
analysis of the compressor is presented in the next section.
See also section 7.7.3 for results of modelling which show the probable inlet valve
trajectory and scavenging mass flows.
-60
-40
-20
0
20
40
60
0
1
2
3
4
5
6
0 0.005 0.01 0.015 0.02 0.025 0.03po
sitio
n (m
m)
Pres
sure
(bar
)
time (s)
mover position
cylinder pressure exhaust valves open
compressor pressure
inlet valves open exhaust valves
close
Chapter 2 Pempek free-piston engine – details and experimental results 30
Figure 2-10 shows a comparison of measured cylinder pressure during the
scavenging period with and without the exhaust pipe attached. All engine settings
were otherwise the same. Apart from a slightly higher initial cylinder blowdown
pressure in the no-pipe case (which is probably just coincidence) the only significant
difference without the pipe is that the cylinder pressure is almost steady. This
comparison illustrates the effects of gas dynamics in an exhaust pipe, even though in
this case the net effect on cylinder charging seemed negligible.
Figure 2-10 Comparison of cylinder pressure with and without exhaust pipe
0
1
2
3
4
0.01 0.015 0.02 0.025
Pres
sure
(bar
)
time (s)
With exhaust pipe
0
1
2
3
4
0.01 0.015 0.02 0.025
Pres
sure
(bar
)
time (s)
Without exhaust pipe
Chapter 2 Pempek free-piston engine – details and experimental results 31
2.4 Compressor analysis
The integrated compressor and passive inlet valves are unique features of the
Pempek engine. Compressor characteristics (swept volume, clearance volume) can
be chosen by design, and though a high compression ratio ensured powerful
scavenging flows, the energy consumption was correspondingly increased. An
indicator plot of the compressor is shown in Figure 2-11.
Figure 2-11 Indicator plot of compressor
The work required for one compressor cycle here is about 48 joules, equivalent to a
cylinder IMEP of 1.35 bar, or about 3kW for both compressors. This is an enormous
loss of energy (almost 30% of the indicated work in the combustion cylinder).
Furthermore, the pumping work is dissipated in the inlet air stream, so assuming
around 440 mg of air is supplied each cycle (see Figure 7-80 below), the supplied air
will be heated above ambient by perhaps 100 °C (not counting heat transfer).
Clearly a reduction in compressor work was necessary.
A modified passive inlet valve mechanism was proposed by Pempek. This design,
as sketched in Figure 2-12 used a counterweight to counteract the natural inertia of
the valves, forcing them to open even with no gas pressure across them. This would
allow the compressor to operate at much lower pressure, while still providing
adequate air delivery. The valves are also tilted to induce strong combustion
chamber swirl. This mechanism has not, however, been tested as yet.
0
0.5
1
1.5
2
2.5
3
3.5
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
Pres
sure
(bar
)
position (mm)
-48J
Chapter 2 Pempek free-piston engine – details and experimental results 32
Figure 2-12 Proposed advanced passive inlet valve design [96]
A further concern for the original passive inlet valve configuration is the relatively
high seating velocity. Modelling suggests the valves bounce several times each
cycle, with seating velocities up to 3.5m/s (see Figure 7-79 below and [55]).
Chapter 2 Pempek free-piston engine – details and experimental results 33
2.5 Summary
The most outstanding success of the Pempek free-piston engine project has been to
demonstrate robust and accurate piston motion control. The control method depends
on a controllable generator and is based on targeting a certain mover velocity -
which if met, will produce a predictable kinetic energy for compression. The control
algorithm can be easily made more sophisticated to suit the efficiency requirements
of the electric machine, for example, to make the typical force profile relatively flat
throughout the stroke. At the same time, it is able to instantly respond to unexpected
changes in combustion energy and keep the mover stroking to its prescribed
compression. A further useful capability of this control method is that it could easily
adjust the running frequency of the engine, which would be necessary if several
modules were run together to achieve dynamic balance. To re-iterate a point made
above in section 1.3, free-piston researchers should consider the demonstrated
control advantage of using a controllable load (be it a hydraulic pump or electric
machine).
On other points, the Pempek engine has been less successful, though this is hardly
unusual for such an ambitious mechanical design project, as the scattered literature
on free-piston engine design testifies (see Appendix I). Amongst all the engine
components developed specifically for this project, the linear electric generator and
electro-pneumatic valves actuators continue to challenge researchers around the
world.
The work reported in this thesis focused in the thermodynamic aspects of the engine,
specifically issues that effected power and efficiency.
Power – Early engine tests revealed substantially lower combustion energy than had
been expected. Conversion of the inlet valve system to four valves instead of one,
and using compression ignition instead of spark ignition improved combustion
energy somewhat, but it was still below expectation. In the ideal case (pure air at
room temperature and pressure), the cylinder would hold 430mg of air which would
should burn about 25mg of diesel fuel. Analysis of the engine’s actual fuelling limit
found that 14mg of fuel was about the maximum that could be efficiently burned.
Early analysis identified several factors that would reduce trapped air mass:
trapped residual exhaust (probably about 8%)
inlet charge heating in compressor (about 100 K)
late exhaust valve closing
Chapter 2 Pempek free-piston engine – details and experimental results 34
late inlet valve closing (this was hard to confirm since the inlet valve
trajectory was unknown)
low initial cylinder pressure (this was hard to confirm since the cylinder
pressure sensors could not measure absolute pressure)
Taken together, these effects could easily account for the apparently poor charging
efficiency of the engine. Initial work focussed on the scavenging efficiency by
attempting to model the detailed in-cylinder scavenging flows with 3D CFD (see
animation titled “points” in Appendix XV for example). However, it was soon
realised that a more comprehensive gas dynamics analysis would be required to
investigate both the passive inlet valve trajectory, and the final cylinder pressure at
valve closing.
Efficiency – The high parasitic load of the integrated compressor was clearly
unacceptable, but some level of inlet pressurisation was probably necessary to drive
the scavenging process. Here again, a gas dynamic model of the engine was deemed
necessary to explore the lower limits of inlet charge compression, and the effect of
exhaust pipe gas dynamics on the charging process.
2.5.1 Motivation for the work presented in this thesis
Thus, this author deemed as crucial a comprehensive gas dynamic engine model for
investigating design options for improving power and efficiency. The model that has
been developed enables simulation of a complete engine cycle in around 2 minutes
on a desktop computer. The designer can use it to investigate:
optimal valve timing
optimal valve sizing
optimal inlet and exhaust duct shape
delivery ratio
necessary inlet charge compression etc.
The model is also useful for investigating piston motion and control issues which can
help the designer to correctly specify the electric machine (or hydraulic pump in the
case of a hydraulic free-piston engine), predict operating speed, evaluate control
methods etc.
The work presented here does not address detailed combustion physics, as indeed
this is an enormous field and has been investigated to some extent already by other
free-piston engine researchers (see section 1.5.3). Even though the piston trajectory
Chapter 2 Pempek free-piston engine – details and experimental results 35
of a free-piston differs compared to conventional cranked engines, there is no reason
in principle why the many recent developments in combustion technology cannot be
equally applied to free-piston combustion.
2.5.2 Synopsis of remaining chapters
The next three chapters describe the important features of the engine model. Chapter
3 describes the cylinder model, which is a single zone thermodynamic control
volume. The model includes a chemical equilibrium calculation for modelling
combustion composition. Chapter 4 describes the gas dynamic model, which is
based on an existing technique, but with several modifications. Chapter 5 describes
miscellaneous other parts of the engine model which were not covered in the
previous two chapters. Chapter 6 explains the integration of all of the sub models
into a complete engine model structure.
Chapter 7 compares the gas dynamics model to a number of experimental test cases,
including a superficial comparison to measured data from the Pempek engine.
Chapter 8 serves as an application example of gas dynamic modelling in free-piston
engines. Firstly, using the completed gas dynamics engine model, the possibility of
lowering the compressor pressure of the Pempek engine is assessed. Secondly, a
radical modification to the Pempek engine is proposed and the model is used to
explore this possibility.
The final chapter summarises and concludes.
36
. . . .
37
Chapter 3 Thermodynamic and gas property models
This chapter details three key models that are applied to the engine modelling
problem.
Single zone thermodynamic cylinder model
Gas property model
Chemical equilibrium model
Models of engine cylinders range from single homogenous control volumes, to
multi-zone models and even to detailed CFD models with detailed chemistry.
However, since combustion rate modelling and emissions modelling are not part of
this study, a simpler single zone model has been adopted.
The cylinder model is carefully implemented to account for the properties of reacting
gas mixtures. Fuel oxidation rate must be prescribed, and can be set using an
appropriate empirical function. The cylinder model also allows mass transfer of
arbitrary mixtures such as fuel injection and piston blowby.
The second model detailed in this chapter is a gas property model which accurately
determines the relevant thermodynamic properties of a gas mixture based on
temperature and composition. This level of detail is necessary since gas properties
bear a significant influence on the thermodynamic model result. Furthermore,
accurate gas properties are also necessary for gas dynamic modelling (see Chapter 4)
The third model described here is a chemical equilibrium model. This models the
progress of chemical reactions during the burning and high temperature phase of the
engine cycle. This approach to modelling chemical heat release is advantageous for
a number of reasons: it ensures that changes in composition are accurately
represented; it automatically evaluates the heat release due to combustion for
arbitrary fuels and mixtures; and it models some important subtleties of high
temperature mixtures such as dissociation.
Chapter 3 Thermodynamic and gas property models 38
3.1 Thermodynamic control volume model
3.1.1 Energy equation for a control volume
The Law of the conservation of energy in a control volume in rate form can be
written as
(3-1)
where is the total enthalpy flow rate of any mass flows into or out of the
volume and will be written as henceforth. The equation can be expanded to
where the rate of work is due to change in volume. For a single species, the energy
equation can be written as
where is the total species mass change rate, including flows and chemical
reactions. By definition, the specific heat of the (ideal gas) species at constant
volume is . For a control volume containing several species, the equation
then becomes
Defining frozen mixture specific heat as
and noting that
and re-arranging yields
(3-2)
Chapter 3 Thermodynamic and gas property models 39
where the total enthalpy of any given flow is
(3-3)
where is the mass flow rate of each species in a flow across the system
boundary
For an ideal gas
Equation (3-2) describes an open control volume containing any number of reacting
chemical species, which is undergoing a change in volume with heat transfer.
Within the enthalpy term , only mass crossing the volume boundary is counted.
However the term for each species in the problem includes species mass
change due to both flows and chemical reactions. If chemical reactions are involved,
the values of species internal energy must be based on an absolute enthalpy scale
so that the relative enthalpies of formation of the various species are correctly
represented. See section 3.2 for more information on evaluating internal energies.
See Appendix III for a further discussion on using the energy equation in engine
modelling and pressure data analysis.
Note that the frozen specific heat is defined for fixed composition. See Appendix
II for a short discussion on this issue.
3.1.2 Numerical solution of the energy equation for a control volume
The most general and convenient description of the energy equation is given by
equation (3-2) since this places no requirements on the control volume to be non-
reacting or even to contain only ideal gas. Figure 3-1 illustrates the numerical
solution of this differential equation in T. The temperature and its rate of change are
known at time t1. An initial guess at the temperature at t2 T2a is made, after which
equation (3-2) is employed to evaluate dT2a/dt. Assuming a parabolic curve in T
between t1 and t2, a refined estimation of temperature T2b is found by
(3-4)
Chapter 3 Thermodynamic and gas property models 40
Figure 3-1 Numerical solution of the energy equation
The process of successively refining the estimation of T2 is repeated several times.
Since it is done in the context of an engine simulation, all the other terms in equation
(3-2) are also iteratively updated as part of the process. The sequence of operations
is described in Table 3-1. The initial guess is based on extrapolated data from the
previous time step. A conservative initial guess for T based on the assumption of
zero rate of change at the current timestep was used to guard against overshoot.
Computationally intensive evaluations such as combustion rate modelling, chemical
reactions, gas property calculations, heat transfer and piston motion were kept to a
minimum and only iterated twice. The second evaluation of pressure was averaged
with the first to guard against oscillatory results in the case of small control volumes
with large pressure dependant mass flows. Piston motion was only evaluated twice,
since the time scale of piston motion is typically much longer than pressure changes
in the cylinder, with the result that piston motion is not likely to deviate much from
the initial estimates. In the case of a conventional cranked engine, piston motion
may be considered fixed by crank-shaft geometry and therefore prescribed at the
beginning of the cylinder calculation with no further iteration necessary. In Table 3-
1, actions that are single spaced do not have to be carried out in order, while actions
listed after a space are dependent on the previous calculation or calculations.
t2 t1
T2a
T1
T2b
Time
Tem
pera
ture
Chapter 3 Thermodynamic and gas property models 41
Table 3-1 Thermodynamic cylinder model calculation sequence
Action Variables Iteration Guess piston motion (1) Initial guess at (conservative extrapolation) Initial guess at mass and species change
Update Set to the same as prev step Set to same as prev step
Evaluate (1)
Evaluate (1)
Update (1)
Evaluate piston motion (2)
Evaluate any combustion (1)
Evaluate any chemical reactions (1) Evaluate any mass flows (blowby, fuel injection) (1)
Evaluate mixture properties (1) Evaluate any heat transfer (1)
Evaluate (2)
Evaluate (2)
Update (2)
Evaluate any combustion (2)
Evaluate any chemical reactions (2) Evaluate any mass flows (blowby, fuel injection) (2)
Evaluate mixture properties (2) Evaluate any heat transfer (2)
Evaluate (3)
Evaluate (3)
Update (3)
Evaluate piston motion (3)
Chapter 3 Thermodynamic and gas property models 42
3.1.3 Validation of the thermodynamic model
A simplified test case was modelled using the solution of equation (3-2) as described
in section 3.1.2. The test case was a closed volume undergoing a sinusoidal
compression-expansion cycle with a compression ratio of 20:1. The contents of the
cylinder was an ideal gas with a constant ratio of specific heats and an initial
pressure and temperature of 1bar and 300K. The modelled pressure was compared
to the analytical solution for isentropic compression of a calorically perfect gas.
Figure 3-2 shows the exact pressure and the relative model error over one cycle for
the case of 200, 100 and 50 time steps. More timesteps improves the accuracy of the
model, but even with a relatively coarse number of time steps, the maximum error is
0.5%.
Figure 3-2 Pressure history and error for various time steps
A more complex test case was modelled which included a combustion event. The
same compression-expansion cycle and initial conditions were used as the above
case, but the initial cylinder contents were 21 parts Oxygen, 79 parts Nitrogen, 1 part
water vapour and 0.038 parts Carbon Dioxide, and 5 parts iso-octane by mass. The
specific heats, internal energy and chemical reaction rates of the gas mixture were
calculated using the methods described below in sections 3.2 and 3.3. Figure 3-3
shows the prescribed fuel burn rate, with ignition suddenly beginning at TDC, and
falling away asymptotically as the fuel is consumed. The fuel burn rate is given in
mass of fuel per mass of cylinder contents per second, and the total time for the cycle
was 20ms. A discontinuous fuel burn profile was selected to provide a severe test of
the thermodynamic model.
0
10
20
30
40
50
60
70
0 0.2 0.4 0.6 0.8 1
P (b
ar)
compression - expansion cycle
-0.6%-0.5%-0.4%-0.3%-0.2%-0.1%0.0%0.1%0.2%
0 0.2 0.4 0.6 0.8 1compression - expansion cycle
error
20010050
Chapter 3 Thermodynamic and gas property models 43
Figure 3-4 shows the pressure and temperature produced by the model for the case of
400, 200 and 100 timesteps.
Figure 3-3 Prescribed fuel burn rate and cylinder volume
Figure 3-4 Pressure and temperature for combustion case
Since an analytical solution for combustion was not available, the simulation using
400 time steps was used as a base line and the other two simulation runs with fewer
0
0.25
0.5
0.75
1
0
10
20
30
40
50
60
70
80
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
vol
fuel
bur
n ra
te k
g/kg
/s
compression - expansion cycle
df_dt
V
0.E+00
2.E+06
4.E+06
6.E+06
8.E+06
1.E+07
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
P (Pa)
compression - expansion cycle
P 400P 200P 100
0
500
1000
1500
2000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
T (K)
compression - expansion cycle
T 400T 200T 100
Chapter 3 Thermodynamic and gas property models 44
time steps were compered. Figure 3-5 shows the relative error in pressure or
temperature for the case of 200 and 100 time steps. The error is larger than in the no
combustion case and this is probably due to the discontinuous nature of the
combustion even. Even in the case of only 100 time steps, the error is mostly less
than 1%.
Figure 3-5 Error for various time steps
It is worth noting that much larger errors may occur if mixture properties are
incorrectly modelled (see Figure 3-7 below). The gas property model described in
the next section was developed in order to ensure that gas property errors are small.
-1.5%
-1.0%
-0.5%
0.0%
0.5%
1.0%
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
erro
r
compression - expansion cycle
200
100
Chapter 3 Thermodynamic and gas property models 45
3.2 Gas mixture property model
Very simple thermodynamic models sometimes use the specific heats of air at room
temperature. However the assumption of constant specific heats may lead to large
errors in engine simulations, since the specific heat of engine gases changes with
temperature, as can be seen in Figure 3-6. Moreover, the specific heat of any given
mixture of gasses will depend on the relative quantities of each gas in the mixture.
Figure 3-6 Specific heat Cp of common exhaust gas species
As a test case, a closed volume undergoing a sinusoidal compression-expansion
cycle with a compression ratio of 20:1 was modelled for both constant specific heats
( =1.4) and for the specific heats calculated with a typical atmospheric mixture of
gases (as described below in section 3.2.2). The constant specific heat case over-
predicted the peak pressure by about 7%, as shown in Figure 3-7.
Figure 3-7 Typical pressure error incurred for setting =1.4
0
500
1000
1500
2000
2500
3000
0 500 1000 1500 2000 2500 3000
0%
2%
4%
6%
8%
0 0.2 0.4 0.6 0.8 1
erro
r
compression - expansion cycle
H2O
N2
CO2
O2
T (K)
Cp (
J/kg
/K)
Chapter 3 Thermodynamic and gas property models 46
A simple thermodynamic model can be significantly improved by dividing the
cylinder process into periods of compression, combustion and expansion and
assigning to each a ratio of specific heats or polytropic exponent that is constant
or linear in temperature [62]. This method tries to account for the variation in gas
properties that occur throughout the engine cycle, and it can also be tuned to allow
the effects of heat transfer and gas leakage to be crudely but effectively modelled.
The best value for is determined based on experience, so this model is not well
suited to predictive simulations.
Krieger and Boorman [72] introduced a model of using mathematical expressions
which are functions of temperature, pressure and equivalence ratio and are valid for
any fuel with a composition of the form . Klein [71] explored various models
for in an engine cycle simulation against a full chemical equilibrium model and
found that while a linear function in temperature is sufficiently accurate for
unburned gas up to a moderate temperature, a polynomial expression such as by
Krieger and Boorman is necessary for burned gasses. Klein proposes a two zone
model consisting of unburned fuel-air mixture and burned gas. Mixtures involving
fuels with differing ratio (to the curves derived by Krieger and Boorman) will
need to be analysed for chemical equilibrium properties and have new sets of curves
fitted. To illustrate the typical variation in , the ratio of specific heats for air-fuel
mixture and for combustion products is shown in Figure 3-8.
Rather than conceiving of engine mixtures as either burned or unburned, better
generality can be achieved with a small increase in computational burden by
considering all of the significant gas species, and this is the approach taken here.
Using this approach, fuels of arbitrary composition can be specified, as can differing
air humidity, EGR, equivalence ratio etc. Also importantly, chemical reactions such
as dissociation, combustion and exhaust after-treatment may be modelled.
Chapter 3 Thermodynamic and gas property models 47
Figure 3-8 Variation of with temperature and equivalence ratio for unburned and burned mixture 1
3.2.1 Species properties
The gas properties of typical products of combustion are taken from the JANAF
thermochemical tables [38] which list the enthalpy and specific heat on a per-
mole basis. These are listed in Appendix IV in Table IV-1 and Table IV-2.
Use of tabulated data may be more computationally efficient than using polynomial
curve fits, since each temperature increment can be made to correspond to an array
index from 1-60. Values are then simply interpolated linearly between the two
nearest temperature values.
Properties of various common fuels are listed in Appendix IV in Table IV-3.
The gas species are presumed to be ideal gases, so the internal energy is found by
1 is the ratio of frozen specific heats – see Appendix II for a short discussion
1.2
1.22
1.24
1.26
1.28
1.3
1.32
1.34
1.36
1.38
1.4
250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000
1.2
1.22
1.24
1.26
1.28
1.3
1.32
1.34
1.36
1.38
1.4
250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000
g*
Unburned mixture by equivalence ratio
0
T (K) Mixture made from 75 parts N2, 15 parts O2 and various parts n-octane (C8H18) by mass
g*
Burned mixture by equivalence ratio
0.2 0.4 0.6 0.8 1
0 0.2
0.4 0.6
0.8 1
100 bar
1 bar
Chapter 3 Thermodynamic and gas property models 48
Where the universal gas constant (J/mol/K)
The specific heat at constant volume is found by
The energy properties can be converted from per-mole to per-kilogram quantities by
dividing by the species molar mass M (g/mol).
3.2.2 Mixture properties
Mixture properties are calculated as
where the mixture contains any number of chemical species s, and is the mass of
each species. The ratio of specific heats for the mixture is
The mixture’s gas constant is
The starred notation is used here to indicate that these are frozen mixture quantities
and are not in general equal to the true specific heats (when shifting chemical
equilibrium is allowed to cause changes to species fractions). See Appendix II for a
short discussion on this issue.
Chapter 3 Thermodynamic and gas property models 49
3.3 Reacting gas mixture model
At low temperatures, mixtures of gases typically do not react so any initially
specified composition will remain constant and the mixture properties can be
calculated directly as described above in section 3.2.2. However at elevated
temperature, the initial mixture composition may change due to chemical reactions.
Combustion reactions at low temperatures (T<1400) can be evaluated analytically,
however at higher temperatures significant extra species may be produced and must
be solved numerically. The importance of these extra species is twofold. Firstly, if
present in large enough fraction, they will affect the energy balance of the mixture,
altering the mixture enthalpy and specific heat. Secondly, even though some species
may not be in high enough fraction to significantly alter the energy properties of the
mixture they may nevertheless be important species to consider as pollutants.
After the onset of combustion, the temperature is usually high enough and reaction
rates fast enough that gas mixtures can be assumed to be in shifting chemical
equilibrium [62]. An exception is the oxides of nitrogen.
3.3.1 Chemical equilibrium model
The model used here uses the concept of equilibrium constants. This method for
calculating equilibrium composition is commonly used [46, 94, 99, 104, 106, 127]
and helpful explanations can be found in [30, 49]
Species considered in the equilibrium calculation are listed in Table 3-2. The
species in this list are the only species to exist in significant quantity at common air-
fuel ratios, temperatures and pressures found in engines. Indeed in normal
circumstances N could probably be neglected since it is rarely present above 1 part
per million. It is included here because it is involved in the reactions that produce
and consume NO. Note also that in very rich mixtures (carbon:oxygen atom ratios
approaching 1), a large number of carbon based species will form.
Chapter 3 Thermodynamic and gas property models 50
Table 3-2 List of species considered in equilibrium calculation
Name Chemical formula carbon monoxide CO carbon dioxide CO2 hydrogen, monatomic H hydrogen, diatomic H2 water vapour H2O nitrogen, monatomic N nitrogen, diatomic N2 nitric oxide NO oxygen, monatomic O oxygen, diatomic O2 hydroxyl OH
The mixture contains a set number of atoms. We can write four mass balance
equations for moles of atomic carbon, hydrogen, nitrogen and oxygen.
Where is the mole fraction of each species and is the total moles of the
mixture. Thus there are 12 unknowns - 11 species mole fractions and the unknown
total moles. Another mass balance equation can be written for the sum of all species
mole fractions, which by definition is 1.
More equations are needed to solve the 12 unknowns. Table 3-3 contains a list of 8
suitable equilibrium equations. The thermodynamic property (Gibbs free energy)
is defined as
where is the molar specific entropy at the reference pressure (0.1Mpa) and is
the absolute enthalpy. P is in bar.
Equilibrium equations 1,2,4,5,6,7,8 were used, along with the five mass balance
equations. Details of the numerical solution that was devised are given in Appendix
V.
Chapter 3 Thermodynamic and gas property models 51
Table 3-3 Equilibrium equations
Reaction Equilibrium equation Equilibrium constant
1
2
3
4
5
6
7
8
To speed up equilibrium computations, the equilibrium constant for each of the eight
reactions was pre-calculated for temperatures between 100 and 6000K. The values
are listed in Appendix IV in Table IV-5
3.3.2 Validation of the equilibrium model
The equilibrium model was compared to the industry standard NASA code CEA
[79] for a range of test cases. Figure 3-9 shows the equilibrium mass fractions
resulting from a set mixture for varying temperature and pressure. Figure 3-10
shows the equilibrium fractions at fixed temperature and pressure for varying fuel to
oxygen ratio. The mixture chosen for analysis is somewhat arbitrary; however it is
representative of the conditions found in an engine. The model corresponds well to
the NASA code results, even though the NASA code considers a comprehensive list
of species, while the model here is restricted to 11 specific species. Potentially
significant mass fractions of NO2 are apparent in the NASA results on the lean side
of Figure 3-10, and ammonia and several carbon based species begin to appear at
very rich mixtures. It is also apparent that at low mole fractions the NASA code
suffers from some kind of rounding error, but this is hardly significant.
Chapter 3 Thermodynamic and gas property models 52
Figure 3-9 Equilibrium species mass fractions of a fuel-air mixture at various temperatures and pressure
Figure 3-10 Equilibrium species mass fractions of a fuel air mixture with varying fuel to oxygen ratio
0.000000
0.000001
0.00001
0.0001
0.001
0.01
0.1
1
1500 2500 3500 4500 55000.000000
0.000001
0.00001
0.0001
0.001
0.01
0.1
1
1 4 16 64 256
0.000000
0.000001
0.00001
0.0001
0.001
0.01
0.1
1
0% 10% 20% 30% 40% 50% 60% 70% 80%
H2
H2O
N2
OH
NO
CO CO2
O H2O
N2
CO2
N
H
N
O2
CO2
CO
NO
OH O2 H2
O
P=1 bar T=3000K
Mixture made from 75 parts N2, 15 parts O2 and 5 parts n-octane (C8H18) by mass
3
dT (K)
C H ) byP (bar)
Mass fraction H
Present code
NASA code Present code
NASA code
H2O
N2
OH
NO
CO
CO2
O H
O2
H2
P=50 bar T=2000 K
NO
OH
CO
H2
H2O
CO2
n-octane (C8H18) to oxygen (O2 ) ratio (kg/kg)
Mixture made from 75 parts N2, 15 parts O2 and varying parts nOctane (C8H18) by mass
Mass fraction
O2
NO2
CH4
HCN
NH3
Present code
NASA code
Chapter 3 Thermodynamic and gas property models 53
3.3.3 NO Rate limit
Due to the importance of NO as an atmospheric pollutant, the model includes a rate
limiting adjustment for NO production/destruction according to the extended
Zeldovich mechanism as described by Borman [30] and Ferguson [49]. This is
necessary because the rates of chemical reactions producing/consuming NO are
significantly slower than the other reactions and the equilibrium assumption is not
accurate in its case.
3.3.4 Notable features of the equilibrium code
Robust for all ratios up to about 0.95. Past this point, the O2 fraction
becomes too small to accurately compute. The only carbon compounds
considered are and , so past this point, excess fuel is retained as un-
reacted fuel in the mixture
Zero hydrogen, carbon and nitrogen are permitted; atomic molar fractions
less than 5x10-6 are rounded down to zero for the sake of numerical
expediency. Only oxygen must be present at a finite fraction.
The code is efficient, reaching adequate accuracy with an average of five iterations
each involving a 4x4 matrix division.
54
. . . .
55
Chapter 4 Unsteady 1D gas dynamics model
4.1 Introduction
Gas dynamic effects on reciprocating IC engines have been recognised for many
years, however gas dynamic effects are not always considered in engine models. A
simple idealised example will suffice to demonstrate the potential modelling errors
from ignoring gas dynamics effects. The mass flow rate out of a reservoir through a
suddenly opened duct is shown in Figure 4-1. The duct is 300mm in length and
connects two reservoirs with a pressure ratio of 1.4. The duct is initially at the
downstream pressure, and is frictionless and adiabatic. This case is similar to
exhaust blowdown, when a valve or port opens rapidly and gas at high pressure
flows into the exhaust manifold.
Figure 4-1 Evolving mass flow rate into an idealised duct
The mass flow from the upstream reservoir takes several milliseconds to establish a
steady value due to the time it takes for pressure changes to propagate to the end of
the duct and back again several times - the longer the duct, the longer the flow
transient. If this flow were modelled as quasi steady (neglecting gas dynamic
effects), the mass flow rate would be greatly over-predicted for the first 2ms or so.
Admittedly, the effects of gas dynamics in real engines are not always as dramatic as
in the example given above. A low speed four stroke engine exhaust stroke may
0
10
20
30
40
50
60
70
80
90
100
0 1 2 3 4 5 6 7 8 9 10ms
Gas properties R=278J/kg/K, =1.4, P1=1.4bar, T1=330.3K, P2=1bar, T2=300K
Duct is 300mm long Results from a gas dynamic model
Mas
s flo
w ra
te %
of s
tead
y flo
w
Chapter 4 Unsteady 1D gas dynamics model 56
take 30ms or more to complete (far longer than the ~2ms taken for the example
above). Often the flow area at the engine valves is somewhat smaller than the
exhaust pipe, and this further militates against gas dynamic influences.
Nevertheless, gas dynamics is known to have significant influence on four strokes
engines, especially high performance engines [95], and both inlet and exhaust
ducting contribute to the ability of the engine to ‘breath’.
Two stroke engines are even more influenced by gas dynamics effects because the
gas exchange period is shorter than a four stroke, large port areas are typical, and gas
exchange is driven by pressure differential across simultaneously open inlet and
exhaust ports, rather than the positive displacement pumping of the four stroke cycle.
A common gas dynamic strategy for increasing trapped mass in two strokes is to
design the exhaust pipe so that a plugging pulse arrives at the exhaust port just
before it closes. Design of exhaust and inlet ducting to maximise the benefits of
natural resonances is notoriously difficult when the engine must operate over a range
of speeds. A compromise is inevitable between high peak power on the one hand
(aggressive tuning for a narrow RPM band), and drivability on the other (a flat
torque curve).
Two stroke free-piston engines such as the Pempek engine are uniquely positioned to
take full advantage of gas dynamic tuning, because they are essentially constant
speed (i.e. frequency) machines. One of the central aims of this modelling project
has been to explore the potential of gas dynamic tuning for the Pempek engine.
4.1.1 Overview of gas dynamic modelling for IC engines
The relationship between pressure and velocity for a pressure wave travelling in one
direction was derived by Samuel Earnshaw [44] in 1860. Bannister and Mucklow
[18] experimentally validated the theory of finite waves using a shock tube
experiment. Riemann [101] had earlier proposed the method of characteristics
(MOC) and this was subsequently developed into a graphical solution for unsteady
flow problems. Benson et al. [22] used digital computing and the MOC to evaluate
unsteady flows in engine ducts. By the 1980’s gas dynamics in engine modelling
using the MOC was well established [21]. Since then finite difference formulations
have become popular for engine gas dynamics [124], though work using the MOC
continues eg [132].
Chapter 4 Unsteady 1D gas dynamics model 57
Despite a long history and the feeling by some that it is time to close the book on
further development in this field [124], diverse, high quality work on unsteady 1D
gas dynamics continues to be reported in the literature. Moreover, there remains an
incredible variety of methods and codes in use in engine research laboratories around
the world that reflect the particular historical needs and emphases of each research
group, and the pedigree of its gas dynamics code, whether it be a MOC based
formulation, other wave action method [24], or a finite difference formulation.
Much of the recent work reported in literature relates to improving boundary or
junction models (eg [19, 29, 36, 54, 92, 111]). Accuracy in modelling boundaries is
critical in engine modelling, since there are numerous duct boundaries that trace the
path from air inlet through the combustion cylinder to exhaust outlet. Furthermore,
new opportunities in detailed CFD present new challenges in interfacing. Since
boundary formulations are typically based on wave action concepts (even finite
difference codes), work reported in this area has relevance for all 1D gas dynamics
codes.
4.1.2 Requirements for the gas dynamics code
A two stroke engine is highly influenced by gas dynamics, and due to the nature of
these engines (sharp flow events, cool air short circuiting, tapered ducts, ports,
valves, crank cases and the like) the code must be good at dealing with these kinds of
flows. A code capable of modelling difficult flows will also excel at modelling less
challenging engines. A code was developed to model the Pempek free-piston engine
with the following requirements in mind.
Mass conservation – not many codes are strictly mass conservative and even
though small discrepancies may be tolerated, any error here will probably
produce an error of the same proportion in engine or turbine/compressor power
predictions [69].
Tapered ducts – tuned two-stroke exhaust pipes usually employ long tapered
sections, so accuracy here is important.
Sudden area changes – There are numerous locations in within the engine’s flow
path where sudden changes in flow area occur. Ports and poppet valves are time
varying flow restrictions.
Gas property discontinuity – Changes in gas temperature and composition within
the passages of the engine have a strong bearing on the passage of pressure
Chapter 4 Unsteady 1D gas dynamics model 58
waves. Regardless of whether the variation in properties is sharp or diffuse, the
effect on pressure waves must be properly accounted for.
Flow losses - Heat transfer in exhaust ducts can be significant, as can the
resulting change in gas temperature. Friction has a small effect in straight ducts,
but can also be used to model separated flow such as in bends and downstream of
restrictions.
Numerical damping – first order discretisation such as is typical of the mesh
MOC tends to smear details in the flow. Compared to a second order scheme and
depending on the wavelength of typical pulsations, more mesh points may be
needed to achieve satisfactory resolution. On the other hand, higher order
schemes must not produce spurious oscillations at sharp changes in gradient.
Computational efficiency – One important advantage of 1D gas dynamics models
over multi-dimensional CFD models is brisk execution. The faster the code, the
more efficiently the engine designer can work through the iterative process of
design and simulation.
Code complexity – this impacts the usability of the code; the ease by which it can
be de-bugged, errors discovered, functionality added or modified and the number
of people who can make use of it.
Moving and deforming ducts – not an essential capability, but moving piston
problems such as the combustion and compressor cylinders of the Pempek engine
are usefully modelled using deformable ducts.
4.1.3 Rational for writing an in-house code
There were a number of commercial engine simulation programs available. The
most prominent were:
Virtual 2/4 stroke – Optimum power [4]
Wave – Ricardo [5]
Boost – AVL [1]
GTPower - Gamma Technologies [2]
Lotus engine simulator - Lotus Engineering [3]
This author decided to write his own gas dynamics code for the following reasons.
Firstly, these packages are designed for conventional cranked engines, utilising cam
driven poppet valves or cylinder ports. It was uncertain how easily they could be
adapted to the unique requirements of the Pempek free piston engine, with its un-
Chapter 4 Unsteady 1D gas dynamics model 59
constrained piston and inlet valves. The integrated compressor also presented
special difficulty, due to its dynamic change of length, and rapid acceleration.
Secondly, there was limited information on the actual simulation methods, and
assumptions used in the commercial programs. They are essentially “black boxes”,
since the average user is not interested in how the results are produced, and
furthermore, the code is proprietary information. Thirdly, commercial licencing was
found to be very expensive. It therefore seemed preferable to maintain control of
this aspect of the engine model.
4.1.4 Summary of the gas dynamics code
The 1D gas dynamics code this author developed is based on the method developed
at Queens University, Belfast eg. [23, 50] and subsequently adopted by the
commercial simulation package Virtual EnginesTM. That method has been re-worked
by this author. The original first order (linear) interpolation of pressure waves
extended to second order and the way heat transfer is incorporated is modified and
includes a method for achieving full mass conservation.
In the model, each computational cell is treated as an idealised constant-area,
constant-property, frictionless duct, so that simple algebraic expression are sufficient
to describe the passage of finite amplitude waves. Area change, gas property change
and friction are accounted for at the interfaces between cells. Connection at duct
ends is treated in exactly the same way. Thus the model uses a uniform theoretical
treatment of boundaries throughout. The boundary solution algorithm summarised
here was developed with careful consideration to the accuracy, numerical efficiency,
and stability of the solution for a wide range of possible flow cases.
This method is similar to the mesh method of characteristics. Common criticisms of
these older wave action methods compared with newer finite difference methods are:
numerical smearing due to first order discretisation, poor mass conservation, and
high computational load. The code presented here addresses all of these issues.
Numerical smearing is reduced by second order discretisation and a method is
introduced to effectively maintain mass conservation. Finally, computational load of
any engine gas dynamics model depends strongly on the number of complex flow
boundaries that need to be solved (which are many). The present method lays great
emphasis on accuracy and efficiency of these boundary flow solutions. The elegance
of finite difference solutions is less obvious when all of the necessary flow
boundaries and gas property variations in a real engine are considered too.
Chapter 4 Unsteady 1D gas dynamics model 60
Validation of the model is described in section 4.6 and Chapter 7. Direct
comparison with other popular numerical schemes is not included here, however
useful comparisons can be found in work by others such as [37, 42, 69, 124].
Chapter 4 Unsteady 1D gas dynamics model 61
Position
Time
Low pressure
Gas in a frictionless duct
Particle path lines
Pressure wave High pressure
4.2 Theoretical Basis
The foundational equation used here for evaluation of unsteady gas flow is that given
by Earnshaw [44] for a pressure wave travelling in one direction where a particle
experiences a change in velocity as a function of the change in pressure.1 The
equation assumes a calorically perfect gas.
(4-1)
Where and are the final and initial velocities, and are the final and initial
pressures, is the ratio of specific heats and is the speed of sound at the original
pressure . The compression or expansion process is assumed to be isentropic, so
the final speed of sound is found as:
(4-2)
In equation (4-1) the positive direction of velocity is in the same direction as the
motion of the pressure wave. The physical meaning of the equation is illustrated in
Figure 4-2. Particles in 1D flow are accelerated toward the right by the influence of
a right travelling pressure wave.
Figure 4-2 A right travelling pressure wave
1 A derivation of equation (4-1) is presented in Appendix VI
Chapter 4 Unsteady 1D gas dynamics model 62
Figure 4-3 Oppositely moving pressure waves
Setting the quiescent velocity in equation (4-1) to zero and evaluating the general
case of oppositely moving pressure waves as shown in Figure 4-3, equations for
velocity and pressure as functions of the left and right travelling pressure waves can
be written as:
(4-3)
(4-4)
Velocity is positive in the rightward direction. The variable is shorthand for
(4-5)
The subscripts R and L signify the rightward and leftward travelling pressure waves
respectively. Where appears without a subscript, this signifies the superposition
pressure – i.e. the static pressure. Note the reference pressure can be set to an
arbitrary value, though it should be close to the typical pressure being modelled. It is
conventional to set it to atmospheric pressure.
P
Position
PR u PL u
u0 P0
Chapter 4 Unsteady 1D gas dynamics model 63
4.3 Wave Propagation
Earnshaw’s equation (4-1) was derived under the conditions of constant flow area,
constant gas properties and no friction or heat transfer. If these conditions are met,
then the waves will propagate through the gas with unchanged magnitude (though
they may distort due to uneven propagation velocities). At any given instant, the
value of the left and right travelling pressure waves at all points along the duct can
be derived from the local fluid velocity and pressure by re-arranging equations (4-3)
and (4-4)
(4-6)
(4-7)
The left and right travelling waves are then advanced on a timestep basis where the
speed of propagation at each point on the wave is the sum of the local speed of sound
and the flow velocity.
(4-8)
Figure 4-4 illustrates the procedure for advancing both waves. The wave speeds
were calculated using the flow velocity and pressure at the diagonally opposite node
along with the mid cell fluid properties. A common timestep is used for all ducts
and volumes in a model. In the engine model, a value of 0.1ms was used.
Figure 4-4 Advancing pressure waves by one time step
The value of the wave incident on the mesh points at the current time step is found
by interpolation of the previous mesh point values. Details are given in the next
section.
time
current time-step
previous time-step
Left travelling wave
Right travelling wave
position
Chapter 4 Unsteady 1D gas dynamics model 64
4.3.1 Second Order Wave Interpolation
In typical mesh MOC solutions this interpolation is linear (or first order) between the
two adjacent mesh points. This has two disadvantages. First, unless the wave
traverses exactly one mesh space in one time step, successive interpolations result in
smearing of features. Second, the distance traversed must not be allowed to exceed
one mesh space, since this would result in extrapolation and solution instability. As
a result the majority of waves in a model will tend to traverse much less than the
ideal one mesh space. The so-called Courant number is the proportion of a mesh
space traversed by a wave in one time step.
To avoid these problems, this author devised a second order wave interpolation
scheme. A third ‘upwind’ mesh point is included in the interpolation procedure, so
that a parabolic curve can be fitted between these points. This improves the
resolution of the model, and permits waves to traverse somewhat more than one
mesh space during a time step. This is illustrated in Figure 4-5.
Figure 4-5 Second order interpolation of pressure waves
Numerical overshoots will appear in the solution near large changes in gradient,
unless precautions are taken. The method adopted here is to simply limit the value
returned from the interpolation to between the upper and lower values in the interval.
If there is any friction, area change or the like at the middle node (which would
modify the value of the incoming pressure wave), then the value of the wave at the
far upwind node must be modified accordingly. Furthermore, if the far upwind node
is outside the duct, then an alternative estimate is required. Details are in Appendix
IX.
x0 x1 x2
Basic curve – three point quadratic
Non-overshoot value limit
position
wave
Chapter 4 Unsteady 1D gas dynamics model 65
4.3.2 Heat Transfer and Mass Conservation
The basic theory of wave propagation outlined above does not allow for heat transfer
to or from the walls of the duct, or indeed for heat released internally due to
chemical reactions. The approach that was developed treats each section of duct
between mesh nodes as a control volume (computational cell). Midway through
each timestep, heat transfer is applied to the gas in each cell, which results in a small
instantaneous change in local pressure, but no change in local velocity. The situation
is illustrated in Figure 4-6. If the cell pressure must be adjusted higher, then both
right and left travelling waves are adjusted higher (by an equal amount), and vice
versa. The steps are as follows:
Figure 4-6 Modifying pressure waves to account for heat transfer and mass
conservation
Step 1 Calculate the mid cell, mid time step pressure before heat transfer is
accounted for.
Step 2 Calculate the mid cell pressure that occurs when heat transfer is accounted
for.
Step 3 Alter the left and right travelling waves according to the change in pressure.
There are two possible ways to estimate the initial mid cell pressure. The first way is
to link the mid cell pressure to the pressure wave values (which are known at the
mesh nodes), and estimate the left and right travelling wave at the cell centre. The
other way is to link the mid cell pressure to the cell mass via the ideal gas relation
. However in the standard implementation of wave action methods, cell
mass is not explicitly linked with local pressure. Thus either mass must be corrected
at each timestep to bring it into line with the local pressure, or the pressure waves
must be modified each timestep to bring the local pressure into line with mass
conservation. The former method is typical - hence those codes cannot be mass
Position x
current time-step
previous time-step
time Right travelling wave Left travelling wave mid
time-step
Chapter 4 Unsteady 1D gas dynamics model 66
conservative, though in practice small accumulating mass error may not be
problematic. The latter method (adjusting pressure waves to conserve mass) is not
desirable either, since this introduces non-physical distortions to the flow solution.
A useful compromise was found that yields good results for most situations. In this
approach mild pressure correction is applied (for any accumulating mass imbalance).
Initial cell pressure is calculated using both methods and but the mass based pressure
value is given heavier weight - in this case 3:1 was used. In this way changes to the
pressure waves are minimised but the scheme retains long term mass conservation.
The method is inevitably a compromise, and provision is made to for the user to
select the non-mass conservative solution if desired.
Step 2 (from above) is to calculate the mid cell pressure due to heat transfer and
chemical reactions. The ideal gas relation is used. Then the
change in magnitude of both left and right travelling waves is calculated as.
(4-9)
4.3.3 Re-meshing
The mesh spacing in a duct should ideally be such that the Courant number is unity.
Practically however, it is impossible to achieve this since any flow velocity will
result in differing wave speed for the left and right travelling waves. Moreover
changes in the temperature and specific heats will also change the wave speeds. Re-
meshing allows the mesh spacing of each duct in a model to be adjusted
independently from time to time to suit changing flow conditions. Re-meshing
replaces existing nodes and cells with new ones where the fluid and flow properties
are interpolated from the existing ones. Figure 4-7 illustrates.
Chapter 4 Unsteady 1D gas dynamics model 67
Figure 4-7 Re-meshing a duct
Re-meshing inevitably introduces some interpolation errors in all flow and fluid
variables, although use of a high order monotonic fitted curve minimises this.
Nevertheless, unnecessary re-meshing is avoided through judicious choice of re-
meshing criteria. The details of the re-meshing criteria are given in Appendix X.
4.3.4 Supersonic flow
Though unlikely in an engine duct, supersonic duct flow is theoretically possible. In
this case the upstream travelling wave will be unable to propagate upstream. The
code will have to test for the presence of a travelling shock, as pictured in Figure 4-
8. Tracing the movement of the shock is not straight forward, though one solution is
to measure the mass in the cell and calculate the shock position to satisfy this. The
speed of the shock relative to the cell can be calculated using the normal shock
equations in Appendix XII.
Figure 4-8 Detection of a travelling shock
time
current time-step
previous time-step
position
Position
current time-step
previous time-step
Left travellingwave XL
L ft t lli
Right travellingwave XR
Super sonic flow
Sub sonic flow
Normal shock
flow
Chapter 4 Unsteady 1D gas dynamics model 68
4.4 Flow Boundary solution
Equation (4-1) assumes constant area, constant property, frictionless, adiabatic flow.
These conditions are too restrictive to directly produce a useful model for the gas
flow in the ducts of real engines. The effects of area changes, friction and changes
in gas properties must be evaluated, and this is done at the mesh points along the
duct. Thus a duct is made up of a string of idealised ducts (cells) where equation (4-
1) and its derivations hold true. At the connection point of each of these idealised
segments, any area change, gas property change or friction is accounted for. The
duct boundaries are solved with the same theoretical treatment, thus a uniform
boundary treatment is used throughout.
It is convenient to name the waves according to whether they are travelling toward
the node (incident) or away (reflected). This is because ordinarily, the flow at a node
is established only by the values of the incident waves (unless the flow is super-
sonic)
Figure 4-9 Duct cell boundary nodes in space and time
In general, to allow for varying gas composition and temperature within a duct, the
gas properties on either side of a node will be different (in space and time). This is
illustrated in Figure 4-10.
Figure 4-10 Variation of gas properties around a node in space and time
Position
current time-step
previous time-step
time
Reflectedwaves
Incidentwaves
Position
time
Reflectedwaves Xr
Incident wavesXi
a Ra a0a b Rb
a0b
c Rc a0c
d Rd a0d
Chapter 4 Unsteady 1D gas dynamics model 69
The flow area may change from one cell to the next. This will occur between the
end cells of two connected ducts of different cross section, and it will also occur
between cells of a tapered duct (since the cell space is assumed to be constant area).
More complex area changes are discussed in the following section.
4.4.1 Flow Types
Considering all possible kinds of flow between duct sections and/or large volumes,1
there are about 31 distinct possibilities as shown pictorially in Figure 4-11. Constant
area duct flow is a special case of type ~1 where the change in area is zero. Friction
or reducing area may cause a choke point such as in type ~2. A sub-sonic diverging
flow will typically experience increased pressure loss due to flow separation. (~3) If
a diverging flow is sonic or super-sonic on the upstream side it will expand further
unless it is decelerated and compressed by a standing shock. (~4-6). A super-sonic
flow may pass entirely through a duct section (~7) unless it becomes choked, or a
shock travelling upstream against the flow passes through the section. In both cases
a travelling shock on the upstream side first slows the flow to a sub-sonic speed
before it passes through section (~8-9). If the flow passes a restricted area between
two ducts, the flow is in two stages. (~10-19). If a duct is connected to a volume
which is supplying the flow, the kinetic energy and pressure of the gas in the volume
are fixed. The inlet flow is assumed to be sub-sonic, though the flow may choke.
(~20-24) If the duct is supplying a flow to a volume, then the pressure on the
downstream side is fixed to the volume pressure (~25, ~28) unless the flow has
choked, in which case the sonic condition fixes the downstream pressure, and the
flow will expand somewhat as it enters the volume (~26, ~27, ~29). If two volumes
are connected directly by a short orifice, the flow may be sub-sonic or it may choke
(~30-31).
1 In the context of a gas dynamic model, a volume is a part of the model where gas dynamic effects are neglected, such as the atmosphere and cylinder. Volumes are typically larger in cross section than ducts, and are modelled as zero dimensional thermodynamic control volumes. A large reservoir such as the atmosphere can be modelled as an infinitely large volume which supplies a steady pressure to any connecting ducts.
Chapter 4 Unsteady 1D gas dynamics model 70
Figure 4-11 Catalogue of all flow types considered
A correct solution for any one of the cases above is not especially difficult - the
difficulty lies in managing the complexity in providing for so many possibilities.
Increasing complexity runs the risk of inadvertent programming error or code
maintenance problems. It is beneficial to arrange the different flow types according
to families and use as much common code as possible.
4.4.2 Solution method overview
The general solution of the flow at boundaries is obtained as a quasi-steady, quasi
1D, calorically perfect flow. These conditions are discussed below.
~1
~2
~3
~4
~5
~6
~7
~8
~9
~10
~11
~12
~13
~14
~15
~16
~17
~18
~19
~20
~21
~22
~23
~24
~25
~26
~27
~28
~29
~30
~31
Single stage flow Two stage flow Volume flow
super-sonic
travelling shock
standing shock
sonic flow
converging flow
diverging flow
stagnant inlet
stagnant outlet
Flow is from left to right
Chapter 4 Unsteady 1D gas dynamics model 71
Steady Flow - The flow at a node (be it a cell or duct boundary) conceptually passes
through an infinitesimal control volume. This allows it to be solved as a steady state
problem since there can be no accumulation of mass or energy within the (small)
control volume. The flow is quasi-steady. It is unclear if there is any alternative to
this approach. Chalet et al. [35] claim to avoid the assumption of steady flow, but
are referring instead to the application of a tuning ‘governor coefficient’ which is a
function of flow Mach number and not based on steady flow-bench data. The quasi
steady assumption remains in place.
1D Flow - The flow is assumed to have uniform properties across any given cross
section. The flow area may change, thus it is not truly 1D but quasi 1D. Clearly
boundary layers, free jets and recirculation zones violate the 1D assumption, but
experimental validation work such as [27, 110] show remarkably faithful simulation
results notwithstanding. The influence of detached flow can be modelled by adding
extra friction at these locations or by including an area coefficient. These methods
are imperfect since they tend not to reproduce the experimental data at all operating
conditions. This author has attempted to model the effect of recent flow history on
flow separation [56]. A modified form of the model is described below in section
5.3.
Calorically Perfect Flow – The quasi steady flow at each node is assumed to have
constant specific heats. Thus the gas is modelled as locally perfect (at each given
instant in time), though specific heats are allowed to vary in both space and time.
The assumption of local calorically perfect flow is not a serious impediment for
engine modelling since there are relatively low pressure ratios (and hence
temperature change) across flow restrictions.
Figure 4-12 shows a typical single stage flow boundary (type ~1). The gas
properties ( , ) in the cell spaces are in general different to the gas properties of
the boundary flow. Conceptually, contact surfaces exist immediately adjacent to the
flow boundary. The pressure and velocity on either side of the flow boundary are
usually different. Unless the flow is isentropic, the downstream isentropic reference
speed of sound increases slightly.
Chapter 4 Unsteady 1D gas dynamics model 72
Figure 4-12 A typical flow boundary showing all flow properties
4.4.3 Example solution
The solution of the flow shown in Figure 4-12 is shown to illustrate the general
procedure. This flow has five unknowns, namely upstream and downstream and
, and downstream . Three equations can be written for the quasi steady flow in
Figure 4-12, between position 1 and 2 (1 2). These are conservation of energy –
equation (4-10), conservation of mass – equation (4-11) and flow energy dissipation,
here described by the change in isentropic reference speed of sound – equation
(4-14).
In addition, two equations can be written for the incident pressure waves as they
traverse the contact surface (a 1 and 2 b). These are equations (4-12) and (4-13).
Note that they are derived from equations (4-6) and (4-7) above.
Energy equation 1 2
(4-10)
Continuity equation 1 2
(4-11)
Wave equation a 1
(4-12)
Wave equation 2 b
a0a a
a0b b
X1 X2 a01 a02
u1 u2
A1 A2
Xia Xib
Unknown values in bold. Flow is from left to right
Chapter 4 Unsteady 1D gas dynamics model 73
(4-13)
equation 1 2
(4-14)
where frictional dissipation may be a function of unknown variables
and the term in the denominator is the should be specified at the pressure of the
friction process (either or both) . Further information on the friction model
used in this thesis is given in section 5.1.
The above five equations form a system of non-linear algebraic equations in five
unknowns which can be solved simultaneously using the Newton-Raphson method
for simultaneous equations.
If the flow on the downstream side exceeds the local sonic velocity, the downstream
wave equation (4-13) must be replaced with the sonic flow equation:
(4-15)
In this case, no information from the downstream side influences the flow.
The derivations of equations (4-10) to (4-15) are shown in Appendix VII.
4.4.4 Separated flow
If the flow is diverging and subsonic (flows ~3, ~12, ~17, ~22 in Figure 4-11), then
it will typically suffer increased pressure loss due to flow separation. This results in
increased downstream thermal energy which is represented by the isentropic
reference speed of sound . In this case a flow separation dissipation term is
added to equation (4-14) so that it becomes
(4-16)
Where is the energy dissipated due to wall friction and is the energy
dissipated due to flow separation. A model for is required. Neglecting wall
friction, classical constant pressure flow can be achieved by setting
Chapter 4 Unsteady 1D gas dynamics model 74
(4-17)
so that the kinetic energy lost from upstream to downstream is entirely converted to
thermal energy instead of pressure recovery. Setting represents isentropic
flow ( ). A model for flow separation has been developed which specifies
to a proportion of the constant pressure (equal pressure) value. See section 5.3
for details.
4.4.5 Two stage flow
In many cases where a duct is joined to another duct or volume, the flow path is
restricted to a smaller cross-section than the flow either side. In this case, the flow
first accelerates toward the restricted throat, and then decelerates into the larger duct
section downstream. To solve this flow two further unknowns must be solved –
namely the velocity and pressure at the throat, and . In this case two flow
equations are added – the energy and continuity equations for the extra stage in the
flow. Note that for simplicity the flow toward the throat is assumed to be isentropic,
while the divergent flow away from the throat may have friction and flow separation
losses applied.
4.4.6 Volume flows
A flow from a volume to a duct differs from duct flow because the cross section of
the upstream flow is undefined and the continuity equation cannot be applied.
However the pressure and kinetic energy of the upstream side are fixed by the
volume state. And so since the upstream and are known, the downstream flow
can be solved using only three equations – energy equation (4-10), the downstream
wave equation (4-13), and the equation (4-14). In many cases, flow from a
volume to a duct should be modelled with a somewhat restricted throat to account for
the vena contracta. In this case, the flow will be two-stage. Note that non-zero
velocity in the volume/reservoir is technically possible, such as modelling the
atmospheric ram air intake on a racing vehicle. In this case, the result is an increase
in dynamic pressure.
The flow from a duct to a volume is a similar problem, however, it is now the
downstream side with fixed pressure and zero velocity. The unknown variables are
the upstream and , and the downstream . The equations necessary are the
Chapter 4 Unsteady 1D gas dynamics model 75
energy equation (4-10), the upstream wave equation(4-12), and the equation (4-
14). A separated jet entering a large volume is generally expected to have no
pressure recovery ( ), so the energy dissipation specified in the equation
should be set to (equation (4-17)). If the flow first contracts to a throat
before entering the volume then it will be two-stage. If the flow velocity exceeds the
local sonic velocity, then the downstream volume pressure no longer controls the
flow. The flow may choke (equation (4-15), or it may be fully supersonic,
depending on the upstream conditions.
4.4.7 Shocks
The effect of a standing shock within a diverging section is automatically satisfied
by the energy, continuity and downstream wave equation, but the possibility of a
fully supersonic outlet must be checked, in which case the downstream wave
equation must be replaced by the equation (4-14).
On the other hand, if the upstream flow is supersonic, checks must be made to see if
the flow will choke, or have a shock migrate upstream from the exit. If the upstream
flow becomes shocked, a simple correction can be made to the upstream incident
wave on each iteration of the main Newton-Raphson solution. This works quite
effectively, since the value of the incident wave changes only slightly when it
traverses such a shock.
4.4.8 Change of reference frame
Ducts with moving boundaries (and by implication, moving internal nodes) can be
modelled by changing the velocity reference frame of the node compared to the cell
reference frame. The change of reference frame modifies the flow velocity (relative
to the computational mesh nodes), but not the pressure. The result is modified wave
values
(4-18)
Where is the velocity of the node relative to the wave’s original reference
frame and positive velocity is the rightward direction. Thus a wave’s effective
magnitude is increased by a node moving to meet it, and decreased if the node is
moving with it. and are the cell space gas properties in which the wave is
travelling.
Chapter 4 Unsteady 1D gas dynamics model 76
Once the flow at the node has been solved, the flow velocity can be converted
back to the cell reference frame simply by
(4-19)
where is the flow velocity relative to the node.
4.4.9 Numerical solution of the flow
Since there are many possible flow geometries (Figure 4-11), as well as sub-sonic
and supersonic versions of each, the first task of the flow solver is to determine
which category of flow the problem belongs to. The direction of a flow through a
given geometry is a crucial factor, so it is important to be able to predict the flow
direction. This is done by calculating the stagnation pressure on both sides of the
flow.
(4-20)
The inlet side will be the side with the highest stagnation pressure.
Once the flow geometry has been analysed and the appropriate equation set
specified, the flow can be solved. In some special cases, the set of equations can be
reduced to a single equation in a single unknown; however most flows require the
simultaneous solution of multiple non-linear equations. The Newton-Raphson
method for simultaneous equations is used here. It requires an ‘initial guess’ for all
of the unknowns, and the solution process is faster and more reliable if the initial
guess is near the solution.
A simple method for obtaining an initial guess of a flow is to assume constant
pressure and constant density across the flow. Then the continuity equation becomes
and for the case of uniform gas properties
Combining these with equations (4-3) and (4-4) and re-arranging gives
(4-21)
and
Chapter 4 Unsteady 1D gas dynamics model 77
(4-22)
where and are the incident waves as shown in Figure 4-10 and Figure 4-12.
Similar approximate solutions can be written for a flow to or from a volume. If there
is a restricted area (throat) between positions 1 and 2, then the velocity at the throat
can be approximated as
If the velocity at the narrowest cross-section exceeds the local sonic velocity, then it
can be clamped to the sonic velocity by solving for the choked flow as
Clearly the ‘constant pressure’ assumption is a significant approximation on real
compressible flows, though it can be quite close for the case of separated diffusing
flows.
It should be noted that an alternative method for obtaining the initial ‘guess’ is to use
the flow solution from the immediately preceding time step. Though this method
has not been used in this thesis, it has the advantage of being computationally
efficient. Nevertheless, some ‘back up’ system is necessary for cases where the
previous timestep flow is very different (such as reversed) or unavailable (such as an
opening valve).
***
Finally, solution of the full set of equations by the Newton-Raphson method is as
follows [98].
Consider the following system of non-linear equations
Then write
Chapter 4 Unsteady 1D gas dynamics model 78
(4-23)
where the partial derivatives and the functions are evaluated at the current
approximation . Equation (4-23) is a system of linear equations with
unknowns which are the correction terms. After solving equation
(4-23), the next approximate solution is.
Equation (4-23) is solved using a built in matrix division function in Matlab [77].
Several iterations of the procedure are required to achieve a sufficiently converged
solution. The convergence criteria used for the flow calculation was for the residual
value of each equation to be below a certain magnitude. These were chosen by
considering the relative magnitudes of typical terms in each equation.
Chapter 4 Unsteady 1D gas dynamics model 79
4.5 Mass and thermal energy transport
The flow of mass and thermal energy can be tracked through the duct system. This
is necessary if any significant variation in gas properties is expected, or if the flow of
certain chemical species is of interest. In engines, it is important to model the
variations in exhaust gas temperature, and this especially the case for two stroke
engines with short circuit flow introducing pulses of low temperature air into the
exhaust duct. In this code conservation variables (mass and isentropic reference
temperature ) are stored in each cell space. This is not the only way to track mass
and heat. An alternative implementation could store only boundary node data.
4.5.1 Cell mass and temperature
Each cell is a control volume with two mass flows and as sketched in Figure
4-13.
Figure 4-13 Mass and thermal transport
If the positive mass flow is defined as flow into the cell, then the cell mass at a new
(mid) timestep can be calculated by assuming the average flow rates since the last
timestep are and .
(4-24)
is the timestep. If the model includes several gas species, then this calculation is
done species by species.
The new cell isentropic reference temperature is approximated as
Position
current time-step
previous time-step
mid time-step
Chapter 4 Unsteady 1D gas dynamics model 80
(4-25)
where the specific heats of the mass flows are assumed constant and identical to the
cell specific heats. Alternatively for true energy conservation, a more rigorous
calculation requiring evaluation of internal energy at would have.
(4-26)
The reference temperature of the new cell could then be back calculated based on
the final cell internal energy . Note that in this thesis, the less rigorous equation
(4-25) was used.
The isentropic reference temperature conveniently uncouples pressure wave energy
from thermal energy, since it is independent of pressure. Foley et al [50] show that
uncoupling wave energy from gas properties avoids cumulative temperature errors.
4.5.2 Boundary flow properties
There are three fluid properties necessary to define an ideal gas for unsteady gas
dynamics - specific heats and , and isentropic reference temperature . Most
other properties of relevance can be derived from these. In this work, the fluid
properties tracked through the model are species mass fraction and reference
temperature . The specific heats are then calculated based on the resulting mixture
composition and temperature.
One simple method to specify the properties of the flow across a cell boundary is to
set them equal to the upstream cell values. This method results in a diffusive flow,
where initially sharp changes in fluid properties are quickly smeared. Some
smearing is physically legitimate since engine duct flow is typically turbulent,
however, the simulated rate of diffusion may be greatly exaggerated.
A model that is capable of preserving discontinuities has been developed bu this
author. It involves constructing a poly-line across the cell upstream of each boundary
flow in each fluid property as illustrated in Figure 4-14. The values of the flow
property at the previous timestep at the upstream and downstream boundary are
and . The cell average of the flow property at the mid time step instant is .
One of two families of curves are used, depending on whether the mid cell value is
closer to the upstream or downstream boundary values. The location of the knee
Chapter 4 Unsteady 1D gas dynamics model 81
in the curve is determined by the criteria that the area under the curve matches that
of the mid cell value . The boundary flow value at the current timestep is then
found as the average value of the curve between and , where is determined
by the estimated flow velocity at the current timestep. Note that the dotted lines on
the lower part of Figure 4-14 represent particle path lines.
Figure 4-14 Calculating boundary flow properties
This solution preserves sharp changes on fluid properties. Extra diffusion can be
added to the solution by moving the point further upstream. A mixing co-
efficient was defined as a fraction of the cell length (0-100%) and this fractional
length was subtracted from the zero mixing position. This method produces
more-or-less consistent mixing effect for both high and low speed flows. In the case
of high speed flows, the time step needs to be small enough to prevent the value
exceeding about 25% of the cell length. Figure 4-15 shows the effect of different
mixing coefficients on an initial temperature discontinuity.
Position
current time-step
previous time-step
mid time-step
0
Flui
d pr
oper
ty
Approximate flow pathline
Flow direction
0
Chapter 4 Unsteady 1D gas dynamics model 82
Figure 4-15 Temperature transport with different mixing coefficients
300
320
340
360
380
400
-10 -5 0 5 10
300
320
340
360
380
400
-10 -5 0 5 10cells
Nodal temperatures after initial discontinuity traversed 20 cells. Flow from left to right
Curves show results for different mixing coefficients
T (K)
T (K)
0 0.1
0.5 1
0 0.1
0.5 1
Flow 300m/s
Flow 30m/s
Chapter 4 Unsteady 1D gas dynamics model 83
4.6 Validation using analytical results
4.6.1 Shock tube problem
The shock tube problem is a useful validation tool for gas dynamics codes because it
contains a shock wave, expansion wave and a contact discontinuity and also has an
exact 1 solution. It presents a severe test of the code’s ability to handle and preserve
sharp discontinuities. Comparisons of various gas dynamics codes using the shock
tube problem can be found in [42, 124].
Table 4-1 Shock tube setup Left Side Right side
(bar) 5 1 (K) 1200 300 (m/s) 0 0 (J/kg/K) 287 287
1.4 1.4 Shock tube length=1m
Diaphragm location=0.5m Computational mesh spacing 0.01m
Time step 0.01ms Results at t=0.5ms
Table 4-1 shows the initial conditions and other settings for the shock tube that was
simulated using the gas dynamics model. Friction and heat transfer were set to zero.
Note that all the shock tube plots below show the nodal (cell boundary) values.
Figure 4-16 shows the results of the model without explicit mass conservation. The
shock front location is well predicted and is resolved over about 2.5 cell spaces and
no overshoot is visible. The expansion wave velocity is slightly over-predicted and
is slightly smeared. The post shock values are all slightly wrong due to the inbuilt
assumption that pressure waves will cause isentropic compression or expansion (as
opposed to the compression process in a shock which is non-isentropic. The
temperature discontinuity is sharply defined and is testament to the transport
property model’s ability to retain sharp discontinuities. (see section 4.5.2 above).
Figure 4-17 shows the pressure and velocity for the same problem but with mass
conservation enforced using the method described above in section 4.5.1. As
expected, this causes some numerical disturbance to the solution with a slight
overshoot apparent around the shock and also just downstream of the expansion
wave peak. The enforced mass conservation is not overly disruptive to the solution,
and interestingly appears to somewhat improve the post shock pressure and velocity.
1 The solution is iterative but converges towards the exact value
Chapter 4 Unsteady 1D gas dynamics model 84
Figure 4-16 Standard shock tube results
Figure 4-18 shows a comparison between the second order wave interpolation
method used in this model, and the linear interpolation method that is commonly
0
1
2
3
4
5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Pressure (bar)
0
100
200
300
400
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Velocity (m/s)
0
500
1000
1500
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Temperature (K)
0
0.5
1
1.5
2
2.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Density (kg/m3)
shock tube position (m)
Mass conservation off
exact model
exact model
exact model
exact model
Chapter 4 Unsteady 1D gas dynamics model 85
used in wave action based codes such as the mesh MOC. The additional smearing of
the first order method is apparent.
It is noteworthy that the shock position is well predicted since no explicit shock
handling routine is used. However, it should be remarked that in any case, this test is
not especially representative of the typical flow in engine ducts, where the pressure
ratios are much lower and shocks usually do not have time to fully develop. In this
context the failure to properly model the non-isentropic compression process across
a shock is unimportant. Nevertheless the code in its present form would be
unsuitable to model problems with very strong shock propagation.
Figure 4-17 Shock tube with mass conservation
0
1
2
3
4
5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Pressure (bar)
-100
0
100
200
300
400
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Velocity (m/s) exact model
exact model
shock tube position (m)
Mass conservation off
Chapter 4 Unsteady 1D gas dynamics model 86
Figure 4-18 Shock tube comparison between first and second order wave
interpolation
4.6.2 Fanno / Rayleigh flow
Two cases of steady flow were compared to exact solutions. Steady flow with
friction and no heat transfer – Fanno flow – was tested by setting a duct with a high
pressure reservoir to the left and a low pressure reservoir to the right. The resulting
simulation was run for a duct with 10 cells till steady flow. The flow at the duct
entry is initially subsonic and high pressure, and as it progresses down the duct it
loses pressure and accelerates until it chokes at the duct outlet. The analytical
solution for Fanno flow based on the downstream boundary flow is plotted for
comparison in Figure 4-19. The results from the gas dynamics model correspond
very well.
Steady flow with heat transfer – Rayleigh flow – was tested in a similar manner. In
this case, friction is set to zero and heat is extracted from each cell at a rate of
6000J/kg/s. The duct has 10 cells. The flow leaves the high pressure upstream
reservoir and enters the duct in a slightly choked state. As heat is drawn from the
flow the pressure increases and the velocity decreases. The analytical solution for
Rayleigh flow based on the downstream duct boundary flow is plotted for
comparison in Figure 4-20. The model drifts slightly from the exact solution but the
error is relatively small. This good result is a significant validation of the heat
transfer implementation (section 4.5.1) since the change in pressure and velocity of
the flow is brought about solely by equal the modification (decrease) of the right and
left travelling waves at each cell at each time step.
2.5
3
3.5
4
4.5
5
0 0.1 0.2 0.3 0.4
Exact
1st order
2nd order
0.5
1
1.5
2
2.5
3
0.72 0.77 0.82
Pressure (bar)
shock tube position (m)
Mass conservation off
Chapter 4 Unsteady 1D gas dynamics model 87
Figure 4-19 Fanno flow
Figure 4-20 Rayleigh flow
The exact solutions for Fanno and Rayleigh flow are shown in Appendix XIII.
320
330
340
350
360
370
1
1.2
1.4
1.6
1.8
2
2.2
0.5 0.6 0.7 0.8 0.9 1
P (bar) T (K)
Exact
P model
T model
1020
1040
1060
1080
1100
0.6
0.7
0.8
0.9
1
0.650.70.750.80.850.90.951
P (bar) T (K)
Exact
P model
T model
Upstream pressure 2.6bar, upstream temperature 394.17K Downstream pressure 1bar
Air, R=287J/kg-K, =1.4 Mass conservation on
0.7 0.8Mach
number
Upstream pressure 1.277bar, upstream temperature 1286.8K Downstream pressure 1bar
Air, R=287J/kg-K, =1.4 Mass conservation on
Mach number
Chapter 4 Unsteady 1D gas dynamics model 88
4.6.3 Numerical smearing
To test the amount of numerical smearing, a high frequency triangular pressure pulse
with a wavelength of 12 cells is introduced into a straight, frictionless, adiabatic duct
and propagates through the duct for 100 mesh spaces. The amplitude of the wave is
small enough that wave distortion due to uneven propagation velocity is negligible.
Figure 4-21 shows the performance of the second order interpolation model
compared to the first order method for different Courant numbers. Both methods
show accumulating interpolation error however there is significantly less smearing
for the second order method which preserves better detail. A phase error is apparent
in the second order model though it should be noted that this smear test pushes the
code’s limits in resolution, and waves with a longer wavelength propagate with
accurate speed.
Clearly, Courant numbers close to unity produce better results, but the first order
method fails spectacularly if the number exceeds one. Compared to linear
interpolation the second order method is markedly improved, with only a small
increase in computational effort.
Figure 4-21 Smearing of a triangular pulse traversing 100 mesh spaces
-90
-60
-30
0
30
60
gau
ge p
resu
re (P
a)
-90
-60
-30
0
30
60
90
gaug
e Pr
essu
re (P
a)
Wave direction 1.01 0.9
0.8 0.4
exact
0.6
Wave direction
1.01
0.9 0.8
0.4
exact
0.6
Cells
Cells
Curves show results for different Courant numbers
Second order wave interpolation With mass conservation
W di ti
First order wave interpolation
Chapter 4 Unsteady 1D gas dynamics model 89
4.7 Summary
The gas dynamic model presented in this chapter has been developed with the
demanding requirements of 2-stroke engine modelling in mind (see section 4.1.2). It
is a wave based method, similar to the MOC, but with improved spatial resolution
due to second order pressure wave discretisation. It is fully ‘non-homentropic’, or
more precisely, allows variable gas properties and temperatures throughout. The
boundary flow solutions allow non-isentropic flows and are fully energy
conservative. The boundary flow solution is common to both duct boundaries and
internal cell boundaries.
Mass and thermal energy are tracked throughout the model, ensuring that gas
properties at various locations are accurately represented. A high resolution mass
and energy transport model allows the user to reliably specify degrees of mixing.
The model is capable of full mass conservation.
The basic gas dynamic model has been validated against a range of analytical test
cases, and shows good performance. The model is further validated against a range
of experimental results below in Chapter 7.
The model can serve as a fully capable basis for modelling internal combustion
engines. However it has not been fully validated against supersonic flows, and some
difficulties here are yet to be resolved. Thus, the model may not be presently
suitable for modelling cases such as supersonic wind tunnel transients or high speed
shock tubes and gun barrels. Similarly, it is not suitable for processes involving very
strong shock fronts, since the compression process is assumed isentropic.
Furthermore, the model assumes the ideal gas relation , so caution is
recommended before this model is applied to potentially non-ideal fluids, such as in
CNG pipe lines.
90
. . . .
91
Chapter 5 Other sub models
This chapter describes the various other sub models that have not yet been addressed
in chapters 3 and 4, but which are significant parts of the overall engine model.
Section 5.1 deals with duct friction and heat transfer together since they are related
phenomena. Even though 1D model allows friction work to be specified (see section
4.4.3) the value is modelled in the present section. Likewise, section 4.3.2 described
the method for heat transfer to be incorporated in the 1D model. The actual value of
the heat transfer must be modelled, and this is the purpose of the present section.
Section 5.2 describes various combustion cylinder sub-models. The Main
thermodynamic volume model described previously in section 3.1 requires inputs for
heat transfer, gas blowby, fuel injection rate and combustion rate. Models for all of
these processes are described here.
Section 5.3 gives the details of a separated flow model introduced previously in
section 4.4.4. Separated flow is a common occurrence in engine ducts, and its
treatment influences the results of the 1D flow model significantly (See section 7.6
below for examples of the significance of the separated flow model)
Section 5.4 deals with the issue of coefficients of flow. It explains the importance of
defining flow coefficients in terms of area, not mass flow. It also describes the way
flow coefficients were implemented in the engine model.
Section 5.5 Describes the multi body dynamics model. This model is necessary for
a free-piston engine, since the trajectories of the pistons and valves are not
prescribed by mechanical linkages.
Chapter 5 Other sub models 92
5.1 Duct friction and heat transfer
Friction and heat transfer in engine ducts are difficult to model due to the complex,
highly unsteady nature of the flows. They are analogous phenomena, both
dependant on the mode of the boundary layer (be it laminar or turbulent). Time
resolved measurements of heat transfer in ducts with unsteady gas flow [20, 130]
have shown that steady correlations are somewhat inaccurate when used to model
instantaneous heat transfer. Most significantly, steady correlations fail to capture the
elevated heat transfer during the turbulent decay period after passage of a flow pulse.
The same problem exists for modelling friction under unsteady conditions. Using
flow over a flat plate as an analogy (see Figure 5-1), the wall shear stress becomes a
function of both velocity and boundary layer development, with a discontinuity
between laminar and turbulent modes.
Figure 5-1 Coefficient of friction for flow over a flat plate [90]
Consider an initially stationary fluid being rapidly impelled by a pressure wave
down a duct for a few milliseconds, before becoming stationary again after the wave
passes. For some initial period of the impulsive flow, the developing flow will be
largely coherent and have a thin boundary layer. It is only after it has travelled some
distance down the duct that the boundary layer thickens and turbulence begins to
form. Conversely, once the flow pulse becomes stationary again, residual turbulent
kinetic energy in the fluid continues to transport heat energy to or from the near wall
area, enhancing heat transfer.
An accurate unsteady friction model would have to model the boundary layer
development of impulsive flow. However, since pipe wall friction is typically not a
0
0.001
0.002
0.003
0.004
0.005
1.E+04 1.E+05 1.E+06 1.E+07 1.E+08
coef
ficie
nt o
f fric
tion
Rex
Turbulent smooth plate
Laminar
Transition
Chapter 5 Other sub models 93
crucial parameter in engine models, there is little discussion in the engine modelling
literature on this issue.
Heat transfer in unsteady conditions is even more complex. Several phenomena are
at work. First, in impulsive flow heat transfer enhancing turbulence will take time
to develop and then once the flow slows, time to dissipate. Secondly, parts of a duct
immediately downstream of some flow restriction will have elevated turbulence
intensity. This turbulent kinetic energy will be convected downstream and will
simultaneously decay. Thirdly, rapid changes in pressure cause compression heating
or expansion cooling of the gas in the thermal boundary layer, and the effect of this
on heat transfer can be significant [20]
5.1.1 Friction model
Provision is made for duct friction in the unsteady gas dynamics model in equation
(4-14) where friction is specified as specific work. The physical situation is
sketched in Figure 5-2. A slug of fluid is passing a computational node and is
dissipating energy through friction, resulting in a flow-wise reduction in pressure
and a corresponding increase in the thermal energy of the fluid. At the instant in
time of the analysis, the slug is centred around the node and has a combined length
of half of the upstream cell and half of the downstream cell.
Figure 5-2 Fluid element experiencing friction
The work per kilogram of fluid passing the node is
fluid element
duct
node
node
node
Chapter 5 Other sub models 94
where F is the force opposing the fluid motion. The wall shear stress can be
expressed as
where is the Fanning friction factor or drag force coefficient and is the fluid
velocity relative to the pipe wall. Thus
(5-1)
where is the duct cross sectional area and is the circumference or wetted
perimeter.
The question of an appropriate friction coefficient was introduced above. A
common correlation for steady flows is the so called Blasius formula for turbulent
flow in smooth pipes with [90]
where the Reynolds number is
And is the hydraulic diameter
The viscosity of air-fuel mixture or exhaust gas can be estimated as a function of
temperature by a fitted power law given in [62]
(5-2)
However, given the intermittent nature of the flows in engine ducts, it is unlikely that
steady flow friction will hold. Based on this author’s analysis of experiments by
Kirkpatrick [68]it was found that a constant friction factor was most suitable as
(5-3)
Chapter 5 Other sub models 95
5.1.2 Heat transfer
The application of heat transfer to the unsteady gas dynamics model is described in
section 4.3.2. The heat transfer is applied to the gas in a cell every (mid) timestep.
The physical situation is sketched in Figure 5-3. Heat is flowing across the walls of
the duct into or out of the gas flowing inside.
Figure 5-3 Fluid element experiencing heat transfer
The total heat transferred over one cell to the fluid in the duct over one timestep is
(5-4)
where is the timestep period. The heat transfer rate is estimated as
where is the fluid temperature, is the duct wall temperature, is the wetted
area and is the heat transfer coefficient.
According to the definition of the Nusselt number, the heat transfer coefficient is
related to the Nusselt number as
Where is the thermal conductivity of the fluid, is the Nusselt number and is
the characteristic length, usually the hydraulic diameter. The thermal conductivity of
the fluid is estimated based on an empirical expression given in [62]
(5-5)
fluid element
duct
node
node
Chapter 5 Other sub models 96
5.1.3 Some other heat transfer models
Blair suggests calculating the instantaneous Nusselt number using the Reynolds
analogy of heat transfer [24] as.
Substituting in the Blasius friction co-efficient yields
(5-6)
Alternatively, substituting in the constant friction factor of equation (5-4) yields
(5-7)
A well-known heat transfer correlation by Dittus and Boelter [43] for steady
turbulent pipe flow is
where n is 0.33 for cooling and 0.4 for heating. Assuming a Prandtl number for the
gas in the duct of 0.77, the Dittus-Boelter correlation becomes approximately
(5-8)
Many researchers base a correlation on the mean duct flow [41, 62]. This method
has the disadvantage of requiring somewhat differing coefficients for differing
engines or duct geometries, and is not suitable for finding instantaneous heat
transfer. Zeng and Assanis [130] propose a two stage model for instantaneous heat
transfer coefficient. In the first stage of an impulsive flow, the model delays the
onset of elevated heat transfer by a correction term based on the instantaneous
acceleration of the flow. Once the flow decelerates to some critical speed, this
correlation no longer holds and a turbulence decay model is employed in this second
stage. Galindo et al [53] describe a model for heat transfer in exhaust ducting that
accounts for the delayed onset of elevated heat transfer during exhaust blowdown,
and also the proximity to the turbulence generating exhaust valve. The model uses a
modified Reynolds number whose velocity term is obtained as a weighted average of
previous calculation instants, and a linear coefficient which corrects for proximity to
the exhaust valve.
Chapter 5 Other sub models 97
5.1.4 Heat transfer model
The model developed by this author is based on the concept of turbulent kinetic
energy in the flow. The model tracks a kinetic energy variable through the flow
and assumes a fixed large eddy length. Sources of turbulence are wall friction and
flow separation (see sections 4.4.3 and 4.4.4). Turbulent kinetic energy is only
monitored for the sake of evaluating heat transfer and does not enter into energy
calculations. The turbulent kinetic energy in a duct cell at each timestep is
calculated as
(5-9)
where the decay term is calculated according to the relationship
where is assumed equal to one and is the large eddy length. Assuming
isotropic turbulence, the turbulent kinetic energy is
then
(5-10)
The large eddy length is assumed to be proportional to the duct hydraulic diameter
The length scale coefficient determines the rate at which the turbulence decays. It
also indirectly controls the magnitude of the turbulence that is allowed to build up in
a duct. The turbulence decay model of Zeng and Assanis [130] used a value of
. However, this causes overly rapid decay of the strong turbulence
introduced downstream of flow restrictions, so the model here uses a much larger
large eddy length of around 20% of the hydraulic diameter.
At each node in the duct, turbulent kinetic energy is generated at the same rate as
friction heating. Thus for the flow at a node, the increase in turbulent kinetic energy
is proportional to the increase in reference temperature.
Chapter 5 Other sub models 98
(5-11)
A correlation for heat transfer coefficient was developed as
(5-12)
where the turbulent Reynolds number is
and
Then the heat transfer coefficient is
(5-13)
This correlation was tuned to mimic the Dittus-Boelter correlation as far as possible
when run to steady state. It assumes that wall friction is calculated according to
equation (5-3) with , and the turbulent kinetic energy is calculated
according to equations (5-9), (5-10) and (5-11).
5.1.5 Some results of the heat transfer model
Figure 5-4 shows the heat transfer model compared to the Dittus-Boelter correlation
for a steady flow thorough a circular pipe. Curves show two different length scale
coefficients. Since the entrance is assumed to be lossless, the turbulence that drives
the heat transfer process takes several pipe diameters to build up to a steady level.
The smaller length scale coefficient (Cl=0.2) causes the model to produce
“developed flow” somewhat sooner than the larger length scale, however the values
converge after about 10 pipe diameters. The Dittus-Boelter correlation assumes
fully developed flow throughout.
Chapter 5 Other sub models 99
Figure 5-4 Heat transfer model compared to Dittus-Boelter for steady flow
Figure 5-5 shows the turbulent kinetic energy calculated by the model for the flow
shown in Figure 5-4. The peak relative turbulence intensity is about 13% and 9.6%
for Cl=0.5 and Cl=0.2 respectively, which is far higher than the real case. This large
discrepancy is the result of the model assuming a constant large eddy length that is
much larger than typically used in modelling turbulence in steady pipe flow. The
large Cl value is used in this model to model the persistence of large, high-energy
turbulent structures caused by separated jet flow through restrictions such as valves.
Figure 5-5 Heat transfer model turbulent kinetic energy for different
turbulence length scales
Figure 5-6 shows the heat transfer for the case of a turbulence generating entrance.
An area coefficient of 0.75 is applied to the flow at the entrance to the pipe, and the
0
50
100
150
200
250
300
350
0 0.05 0.1 0.15 0.2 0.25 0.3
Ch (W
/K/m
2 )
m
Dittus-Boelter
Cl=0.5
Cl=0.2
0
100
200
300
400
0 0.05 0.1 0.15 0.2 0.25 0.3
ket (
J/gk
)
position (m)
Cl=0.5
Cl=0.2
Pipe diameter 25mm, velocity ~135m/s, gas temperature ~400K Flow is from left to right. Entrance assumed lossless
Pipe diameter 25mm, velocity ~135m/s, gas temperature ~400K Flow is from left to right. Entrance assumed lossless
Chapter 5 Other sub models 100
diverging flow downstream of the vena-contracta induces a small total pressure loss,
which is credited to increased turbulent energy. Unlike the ideal entry flow of
Figure 5-4, this case begins with above average heat transfer which decays to a
steady value after 8-10 diameters downstream. Unfortunately since the turbulence
generated at the pipe entry is constant, use of a larger length scale coefficient Cl
tends to reduce the effect of turbulence introduced by separated flows, compared to a
smaller choice of length scale.
Figure 5-7 shows the model compared to the Dittus-Boelter correlation for a slower
flow. Curves show results for two different length scale coefficients.
Figure 5-6 Heat transfer model for a turbulence generating inlet
Figure 5-7 Heat transfer model compared to Dittus-Boelter for low speed flow
0
100
200
300
400
500
600
0 0.05 0.1 0.15 0.2 0.25 0.3
Ch (W
/K/m
2
position (m)
Dittus-Boelter
Cl=0.5
Cl=0.2
0
10
20
30
40
50
60
70
0 0.05 0.1 0.15 0.2 0.25 0.3
Ch (W
/K/m
2
position (m)
Dittus-Boelter
Cl=0.5
Cl=0.2
Pipe diameter 25mm, velocity ~125m/s, gas temperature ~400K Flow is from left to right. Entrance area coefficient 0.75
H t t f d l d t Ditt B lt f l
Pipe diameter 25mm, velocity ~20m/s, gas temperature ~400K Flow is from left to right. Entrance assumed lossless
Chapter 5 Other sub models 101
Figure 5-8 shows results of increasing the number of computational cells from 10 to
20. The results for this case are shown to be grid independent.
Figure 5-8 Heat transfer model for different cell spacing
The steady flow solution is largely unaffected by choice of large eddy length
(providing equations (5-10) and (5-12) both use the same length). Moreover, the
steady solution is not grid or timestep dependant.
To investigate the unsteady behaviour of the model, it was applied to a simulation of
a single shot experiment by Kirkpatrick[68]. Details of the experiment are given in
section 0. A flow pulse enters a pipe from the left and a pressure wave is transmitted
toward the right. Figure 5-9 shows the modelled heat transfer coefficient at two
instants as the pulse travels to the right. Intense turbulence generated by the flow
past the valve at the left enhances the heat transfer in this part of the pipe. The
turbulence is convected downstream but rapidly dissipates. Once the flow pulse
passes, decaying residual turbulence continues to enhance heat transfer. Choice of
length scale affects the rate at which it decays, but does not much effect the peak
heat transfer.
0
10
20
30
40
50
60
70
0 0.05 0.1 0.15 0.2 0.25 0.3
Ch (W
/K/m
2
position (m)
Dittus-Boelter
20 cells
10 cells
Pipe diameter 25mm, velocity ~20m/s, gas temperature ~400K Flow is from left to right. Cl=0.5 Entrance assumed lossless
Chapter 5 Other sub models 102
Figure 5-9 Heat transfer modelled for single shot using different turbulence
length scales
Figure 5-10 shows the same case compared with a timestep of double the size. Note
that the mesh spacing is approximately doubled as well. The results are a close
match which demonstrates that the model is essentially not grid or timestep
dependent.
0
100
200
300
400
500
600
700
800
900
0
20
40
60
80
100
120
140
160
180
0 1 2 3 4 5 6
Ch (W
/K/m
2
velo
city
(m/s
)
position (m)
velocity
Cl=0.2
Cl=0.5
-100
0
100
200
300
400
500
600
-20
0
20
40
60
80
100
120
0 1 2 3 4 5 6
Ch (W
/K/m
2
velo
city
(m/s
)
position (m)
velocity
Cl=0.2
Cl=0.5
Chapter 5 Other sub models 103
Figure 5-10 Heat transfer modelled for single shot using different timestep size
5.1.6 Summary
A reliable, predictive heat transfer model for engine ducts is crucial for modelling
exhaust turbine performance (if one exists), for accurately predicting wave speeds in
engines with highly tuned exhausts, for temperature dependant exhaust after
treatment systems, and so forth. The model developed here is an attempt at a
universally applicable model. Its primary strength is in modelling the turbulence
decay period, which has been found to contribute around half of the total heat
transferred under conditions of intermittent flow [20]. Turbulence enhanced heat
transfer downstream of flow restrictions (ie in regions of separated flow) is naturally
incorporated into the model without having to resort to ad hoc correction terms such
as that described by [53].
The model does not consider the effect of pressure variation on boundary layer
temperature. Though this omission may not have a significant effect on total heat
transfer, it does appear to affect the instantaneous heat transfer rate [20].
Furthermore, the model does not appear to accurately model the delay of enhanced
heat transfer which has been noted by several experimental studies. Heat transfer in
the model is closely tied to flow velocity, at least during acceleration phase. The
model is, for better or worse, dependant on the friction model for turbulence
production. The failure of equation (5-3) to properly model friction for steady flow
pipe is surely cause for concern, but at this stage, a simple but faithful unsteady
friction model has not been developed. A subtler friction model may improve the
delay effect for heat transfer.
-100
0
100
200
300
400
500
600
-20
0
20
40
60
80
100
120
0 1 2 3 4 5 6
Ch (W
/K/m
2
velo
city
(m/s
)
position (m)
velocity
dt=0.0002
dt=0.0001
Chapter 5 Other sub models 104
Other causes for concern are the assumption that all wall friction work is converted
to turbulent kinetic energy, and the arbitrary and unrealistic values for turbulence
that are produced for steady flow (see Figure 5-5). The model has only been
superficially validated ‘by eye’ against the results of Bauer et al [20] and Zeng and
Assanis [130], and also by matching single shot experiments by Kirkpatrick [68] (see
section 7.4). Further work would involve a more thorough theoretical analysis, and
further validations for a range of cases. An improved friction model may be helpful
in further improving the universality of the model.
Chapter 5 Other sub models 105
5.2 Combustion cylinder models
5.2.1 Heat transfer
The instantaneous cylinder heat transfer is estimated using the Annand [14]
correlation where
Where a is constant of proportionality and is set to
as recommended by [49] for 2-stroke engines.
Defining the characteristic length as piston diameter gives
(5-14)
Viscosity is calculated as
(5-2)
Thermal conductivity is calculated as
(5-5)
Since the Pempek engine runs with a nearly constant speed, a mean piston speed of
8m/s was used as the characteristic velocity.
The wall temperature of the combustion chamber was assumed (no wall conduction
or water side heat transfer modelling was done). Three zones were designated -
cylinder head, cylinder wall and piston crown. The instantaneous heat transfer rate
was then estimated as
(5-15)
where T is the average gas temperature in the cylinder. Hot engine combustion
chamber temperatures were assumed as 400K
Chapter 5 Other sub models 106
5.2.2 Blowby
Cylinder pressure measurements of the prototype engine suggested that significant
combustion chamber leakage was occurring (see section 2.3). The likely location of
this leakage was the piston-cylinder crevice, since it had no traditional piston rings.
Instead there was a cylinder mounted ring located far from the piston crown. A
sketch of the piston and cylinder layout is shown in Figure 5-11. The cylinder is
cylindrical, while the piston is slightly tapered to allow for greater thermal expansion
near the crown. The taper angle is 0.019 degrees, and the annular piston-cylinder
gap at the piston crown is 50 microns when cold. The cylinder bore is 68mm.
Figure 5-11 Sketch of piston-cylinder crevice
A CFD model of the problem was created by this author to analyse the likely blowby
flow characteristics. The model was 2D and used 9 cells across the long, narrow
crevice region. Details of the computational mesh are shown in Figure 5-12. The
model was for compressible flow, the density calculated according to the ideal gas
state equation, turbulence model was k- , and heat transfer was calculated based on
fixed wall temperatures. The cylinder pressure was simulated with a deforming
combustion chamber and a heat source term to simulate combustion heat release.
The wall velocities were modelled. The resulting cylinder pressure and crevice flow
are shown in Figure 5-13.
Cylinder wall
piston
labyrinth grooves cylinder ring
Chapter 5 Other sub models 107
Figure 5-12 Blowby CFD model mesh
A simple blowby correlation was developed to allow computationally inexpensive
modelling of the blowby. The blowby correlation is
(5-16)
where
and
for
and
for
The correlation is compared to the CFD value is shown in Figure 5-13. All
quantities in the correlation are in SI units (but note the plot uses bar and
grams/second for convenience). To model the effect of a smaller or larger piston
cylinder gap, the correlation is scaled linearly.
cylinder piston
labyrinth grooves Piston-cylinder crevice
Chapter 5 Other sub models 108
Figure 5-13 Blowby correlation
The model has limited validity. The behaviour in the case of a smaller gap has not
been modelled, and the crevice is assumed to be axisymmetric (i.e. a perfectly
aligned piston). In reality, the piston is probably eccentric, so the gap is much larger
on one side than the other. Furthermore, the piston axis may be tilted with respect to
the cylinder axis, introducing further three dimensional aspects to the problem.
More detailed modelling in three dimensions would be required to properly
understand the characteristics of the blowby flow here.
-20
0
20
40
60
80
100
0 0.005 0.01 0.015 0.02 0.025
bar,
g/s
seconds
P
mass flow CFD
mass flow correlation
Chapter 5 Other sub models 109
5.2.3 Fuel injection
Fuel injection (even of liquid fuels) into the cylinder is assumed to be in the gaseous
state since the cylinder model assumes all species are ideal gases. Thus the process
being modelled is actually fuel evaporation. The rate, timing and total injected mass
are set by the user. The injection (or more accurately evaporation) is assumed to
happen at a constant rate, until the total injected mass specified by the user is
reached.
If the injected fuel is gaseous the molar specific enthalpy is calculated as
(5-17)
where is the enthalpy of formation of the gaseous phase at 298.15K and is
the sensible enthalpy of the gaseous phase. The subscript denotes the temperature
of the injected fuel.
If the fuel is initially liquid, the enthalpy of evaporation and the kinetic energy of the
jet must be accounted for.
(5-18)
where is the enthalpy of evaporation, is the sensible enthalpy of the liquid
fuel, and are the injection pressure and instantaneous cylinder pressure, and
is the liquid fuel density. The subscripts and denote liquid and gaseous
phase and is the boiling point of the fuel.
Figure 5-14 describes how the enthalpy of the liquid fuel is related to the gaseous
phase.
Figure 5-14 Finding fuel injection enthalpy
T
h
liquid phase
gaseous phase
298.15 Tinj
BP
Chapter 5 Other sub models 110
5.2.4 Spark ignition combustion
The combustion rate for spark ignition is modelled with three parameters as shown
in Figure 5-15. The rise time determines the slope of the initial rate rise. After this
period of time, the rate is fixed at a constant value. The burn rate is also limited to a
proportion of the fuel mass still existing. Thus, the burn rate falls asymptotically to
zero as the unburned fuel in the cylinder is depleted.
Figure 5-15 Spark ignition combustion rate model
The parameters are:
Rise time
Maximum rate (normalised by cylinder mass)
Rate coefficient (multiplied by the remaining fuel mass)
time
burn rate kg/s
maximum rate
rise time
unburned fuel mass limited rate
ignition
Chapter 5 Other sub models 111
5.2.5 Compression ignition combustion
Compression ignition is modelled using the fact that fuel-air mixing is finite rate.
The combustion rate is calculated by integrating the fuel mass in the cylinder WRT
time. The burn rate coefficient X is calculated at each time step as
(5-19)
Where is the previous value, is the fuel mass in the cylinder, is the
time step and is the fuel burned since the last time step. The combustion rate
is made proportional to X through use of a user-specified constant so that
(5-20)
The ignition timing is set by the user. The burn rate is related to the fuel injection
timing, as shown in Figure 5-16. Early injection increases the peak combustion rate.
The fuel injection period may overlap with the combustion period. After fuel
injection ceases, the combustion rate falls asymptotically to zero as the unburned
fuel in the cylinder is depleted.
Figure 5-16 Compression ignition combustion rate model
The model is crude, but allows the general combustion behaviour of direct injection
systems to be simulated.
time
burn rate kg/s
end of injection
Fuel mass limited rate
start of injection
ignition
Chapter 5 Other sub models 112
5.3 Separated flow model
Perhaps the most difficult aspect of duct flow to model is the case where the flow
separates from the duct walls and generates energy dissipating turbulent eddies.
Unfortunately, this situation is relatively common in internal combustion engine
ducts. Separated (or diffusing) flow occurs at sudden enlargements in duct area,
turns, throttle valves, cylinder valves or ports and multi-pipe junctions. The problem
for modelling this kind of flow is to determine the level of pressure recovery from
the decelerating flow. It is well known that gently tapered diffusers can decelerate a
flow fairly efficiently, generating high pressure recovery, while steep diffusers are
inefficient, and the pressure recovery may be negligible. A common approach in
fluid mechanics to modelling pressure recovery in stepped ducts is to apply the
momentum equation to a control volume containing the area change, as illustrated in
Figure 5-17.
Figure 5-17 Control volume for applying the momentum equation to a diffusing
flow
Assuming that the pressure on the left hand boundary of the control volume is
uniformly equal to , and that the velocity across the cross section at either end of
the control volume is uniform, and neglecting wall friction, conservation of
momentum can be written as
(5-21)
This approximate solution is used by Blair [24] and has been shown to produce
generally good results [27, 68].
This author has developed an alternative model, based on direct specification of the
change in thermal energy of the flow, according to equation (4-16). (see section
4.4.3 and 4.4.4)
u2 2T02 / 2
A2
P1 Pu1 T01 / 1
A1
P2 Control volume
Chapter 5 Other sub models 113
where is the energy dissipation due to wall friction, and is the energy
dissipation due to flow separation in a diffusing flow. The ability to control the
energy dissipation due to flow separation provides great flexibility. Many different
models can be suggested. It was observed that for a constant pressure flow
( ) the energy dissipation is equal to the loss of kinetic energy
(4-17)
Initially a function was developed that applied a fraction of to the flow
according to the relative deceleration of the flow [56]. Maximum deceleration was
for the case of an infinite expansion in duct area. This model conveniently allowed
both local and convective acceleration to be taken into account. The function is
given in equation (5-22)
(5-22)
However, it was found that the local acceleration term (rate of change of velocity at
a fixed point in the duct) was negligible compared to the convective acceleration
(acceleration of a particle due to its motion through the duct). Furthermore, the
model suffered from both time step and grid dependencies. The effect of
unsteadiness on duct flow separation is still not clear and further work is required
before the details of this complex phenomenon are understood.
In the meantime, a simpler function which does not take into account unsteady
effects and more closely mimics the momentum equation was implemented.
(5-23)
In the case of tapered ducts, equation (5-23) would be mesh dependant since the
ratio of duct areas approaches 1 when the mesh spacing approaches 0. The
correlation can be modified so that it becomes a function of taper slope only. A
sketch of one cell of a tapered duct is shown in Figure 5-18. The slope of the taper is
. If a ‘standard’ cell length is defined as , then the basic function (5-23)
becomes.
(4-16)
Chapter 5 Other sub models 114
(5-24)
This equation is used for tapered ducts. The ‘standard’ cell length used in this thesis
was
Figure 5-18 Modified area ratio for tapered ducts
Equations (5-23) (for sudden area changes) and (5-24) (for tapered ducts) constitute
the separated flow model used in this thesis. Their performance is tested against
experimental results in section 7.6. They are grid and timestep independent, and are
easily applied to the numerical flow solution because they are effectively a constant
coefficient of .
duct centre line
L
l r1 r2 rstd
Chapter 5 Other sub models 115
5.4 Flow area coefficient maps
It has been found that flow through sudden changes in area is often less than would
be predicted if the full cross section was utilised by the flow. One reason for this is
the vena contracta effect. Flow models account for this by including a flow
coefficient of some kind, which is tuned to reproduce experimental results.
Traditionally the flow coefficient or coefficient of discharge is defined as
(5-25)
where is the measured mass flow rate, and is the theoretically obtained
mass flow rate using the full geometric flow area and some form of isentropic nozzle
theory. However, this formulation is unsuitable for this gas dynamic model because
the model requires conservation of mass as a crucial part of the boundary flow model
(see section 4.4.3 above). What is needed therefore is not a mass flow coefficient,
but a flow area coefficient. This coefficient takes the form
(5-26)
where is the flow area which when used in the model reproduces the same
mass flow as the experiment. Determining the flow area coefficient requires an
iterative approach, where the flow model (including the all non-isentropic effects
such as flow separation) is used with successive guesses for until the
experimental pressure ratios and mass flow rate coincide [24]. With the area
coefficient thus defined, the flow area to be use in the model can be found by
multiplying the geometric flow area with the area coefficient.
In the absence of in-house flow data, this author analysed some published flow
coefficient data from QUB [25, 28, 29]. Unfortunately, this Author could not be
certain over the exact theoretical model used to derive the original coefficients, and
this uncertainty was compounded by some apparently inconsistent results.
Flow area coefficients for flow through an array of poppet valves was numerically
obtained using a 3D CFD model, then processing the data as described for equation
(5-26). Figure 5-19 shows the flow area coefficient map for flow through exhaust
valves (from a volume to a duct). The data points from the 3D CFD model are
shown by black circles. The green surface represents the map. The user can visually
modify the surface by dragging points of it up or down. This form for creating and
Chapter 5 Other sub models 116
editing flow area coefficient maps was created using MatLab. Further details of the
area coefficient data structure can be found in Appendix VIII.
Figure 5-19 Example flow area coefficient map
Flow coefficients are a crucial part of a 1D gas dynamics model, as they allow
complex three dimensional flows to be accurately represented in a simplified 1D
world. This author has only built up a limited database of flow area coefficients, and
much of this data is still uncertain. Unfortunately, much existing published data is
difficult to utilise, either because its graphical presentation is not easy to digitise, or
because the exact details of the experiment are unavailable, or because the exact
details of the theoretical flow used to produce the coefficients is either not specified
or difficult to reproduce.
Chapter 5 Other sub models 117
5.5 Multi body dynamics
The Pempek free-piston engine contains several un-constrained parts. These are the
free-piston (or mover), passive inlet valves and electromagnetic exhaust valves.
Figure 5-20 shows these parts in a simplified cutaway sketch.
Figure 5-20 Cutaway of FP3 showing moving parts
5.5.1 Mover dynamics
The forces on the mover are shown schematically in Figure 5-21.
Figure 5-21 Forces on the mover
The magnetic force is applied to the mover by the linear electric machine. This is
the only means of power extraction and also functions as a control on mover motion.
The generator load force is modelled using the same control algorithm as employed
on real engine (see equations (2-1) and (2-2).
Generator force is limited to a certain magnitude that corresponds to the electric
current limit imposed on the machine.
mover passive inlet valves exhaust valves
magnetic force
gas pressure friction
mover
Chapter 5 Other sub models 118
Gas pressure exerts forces on all surfaces of the mover, but only force in the
direction of motion is considered.
Friction includes the mechanical friction of the magnet holder rubbing on the
generator stator surface, as well as electromagnetic effects such as eddy current drag
and magnetic hysteresis. Thus friction is modelled with a constant and variable
component
(5-27)
Where the friction force is opposite to the direction of relative motion, is a
constant, and is the relative motion between the mover and stator.
5.5.2 Exhaust valve dynamics
There were several different exhaust valve actuators used throughout the life of the
project. One of the versions is shown in Figure 5-22. It is made of a magnetically
permeable armature which is connected to an array of four poppet valves with small
springs. The springs ensured that all valves could seat.
Figure 5-22 Forces on the exhaust valves
valves
gas pressure
armature connecting
springs
Chapter 5 Other sub models 119
For the sake of simplicity, the motion of the armature was prescribed (based on
experimental measurements) and the kinetic trajectory of the exhaust valves was
modelled with a combination of gas pressure force, spring force and friction. The
valves could possibly bounce when returning to the seated condition.
Friction from the valve guide was modelled as a constant friction force.
Gas pressure force was modelled by utilising an empirically derived valve force area
coefficient which was a function of both valve lift and pressure ratio as shown in
Figure 5-23. In these maps, the reference area is the circle inscribed by the valve
diameter. AR is the ratio of outer valve curtain area to reference area (or
4*lift/diameter). The data points were taken from a detailed 3D CFD model of the
exhaust valves and inlet valves undergoing steady flow, since physical data on the
aerodynamic forces on the valves was not available.
Figure 5-23 Aero force coefficients for normal flow through exhaust and inlet
valves
5.5.3 Passive inlet valve dynamics
A section view of the piston crown is shown in Figure 5-24, along with the passive
inlet valves (in this diagram there are four).
Normal Exhaust flow Normal inlet flow
Chapter 5 Other sub models 120
Figure 5-24 Forces on the passive inlet valves
The passive inlet valves act under the combined influence of gas pressure forces,
closing spring forces and guide friction. At the same time, the piston on which they
are mounted is often rapidly accelerating. Like the exhaust valves, they may bounce
against their seats.
Gas pressure force was modelled in the same way as for the exhaust valves, using
the aerodynamic force coefficient maps shown in Figure 5-23.
The advanced inertia driven passive inlet valve design sketched above in Figure 2-12
contains a counterweigh mechanism. Although this mechanism has not yet been
modelled or built, the multi-body dynamics code described in the next section is
capable of modelling this kind of complex multi-body dynamic interaction.
5.5.4 Multi-body dynamics model
Bodies in the engine model represent moving parts in the real engine. A body may
be fixed (for example, the engine block), or it may be given a prescribed motion (for
example, a crank driven piston, or cam driven valve.) Alternatively, a body may
move with a kinetic trajectory in response to an array of forces acting on it. The
forces are of two categories. The first category is body interaction forces. These
include spring type forces, friction and impacts or seating forces. All of these forces
act equally and oppositely on a pair of bodies. The second category is any force
which affects only one specific body. Gas pressure force is one such force.
The instantaneous acceleration of a body is
valves
gas pressure
closing springs
Chapter 5 Other sub models 121
(5-28)
The instantaneous velocity is found by assuming acceleration changes linearly
between the previous time step and the current
(5-29)
The instantaneous position is then
(5-30)
where is the time step increment and the subscript ‘prev’ denotes the preceding
time step value. Since some forces acting on a body may themselves depend on the
current position and velocity, the position and velocity are updated (iterated) several
times for each time step increment. The body must be sufficiently massive that its
motion over the period of one timestep is relatively ‘smooth’.
Impacts of one body against another must therefore be dealt with specially. If a
collision condition has been set by the user, and the trajectories of the two bodies are
found to have intersected, then the following procedure is followed:
Calculate the time of impact
Calculate the pre and post impact velocities and impact position
Update any friction forces to account for the reversed relative velocities.
Calculate the final positions and velocities
If the bodies are still intersecting at the end of the time step period, evaluate
them as a single lumped mass
The collision must satisfy conservation of momentum, and a coefficient of restitution
set by the user which specifies the ratio of the relative approach and rebound
velocities. If two bodies become pressed together (such as a seated poppet valve),
then they are treated as a lumped mass. A typical impact involving two bodies with
different mass is shown in Figure 5-25.
Chapter 5 Other sub models 122
Figure 5-25 Typical collision trajectory
This one dimensional multi-body dynamics model allows many of the significant
moving parts of a free-piston engine to be modelled. The carefully implemented
collision module ensures that unconstrained poppet valve motion is accurately
simulated. The dynamics model can safely use the same time step size as the
cylinder and gas dynamics models which are typically set to give between 100 and
400 time steps per piston cycle. Some details of the data structures for bodies and
body interactions can be found in Appendix VIII
Time of impact
time
position
Final time
Initial time
body A
body B
123
Chapter 6 The engine model – integrating all sub
models
This chapter describes the integration of the sub-models detailed in Chapter 3,
Chapter 4 and Chapter 5. The model structure has been deliberately designed to be
multi-purpose. It is capable of modelling devices as simple as a single shock tube, to
as complicated as a multi-cylinder engine. In Chapter 7 below, this multi-purpose
model is applied first to modelling a series of single shot gas dynamics experiments
(by others), and secondly, to modelling the Pempek engine.
In any reciprocating internal combustion engine, there are one or more cylinders.
The flow of air, fuel and combustion products into and out of the cylinder is
controlled by various valves or ports. Usually a duct is interposed between the
cylinder and the outside world. In some cases, there are other volumes, such as
intake plenums, positive displacement compressors (such as crank case scavenged
engines), and mufflers. The application of the various sub models to the various
parts of an IC engine is shown diagrammatically in Figure 6-1
Figure 6-1 Integration of sub models to make an engine model
Thermodynamic control volume model (3.1)
Cylinder process models(5.2) -Heat transfer -Blowby -fuel injection rate -Combustion rate
(spark ignition or compression ignition)
Gas mixture properties model (3.2)
Reacting mixture model (3.3) (chemical equilibrium)
Heat transfer and friction Model (5.1)
ID gas dynamics model (Chapter 4)
Flow coefficient library (5.4)
Body dynamics model (5.5) -piston -exhaust valve inlet valve
separated flow model (5.3)
Engine ducts cylinder
Chapter 6 The engine model – integrating all sub models 124
6.1 Overview of model building blocks
The engine model is composed of three main building blocks. These are:
Ducts – the gas dynamic parts of the model. They are given a length and cross
section, which may be tapered. A duct is made of n cells with n+1 cell boundaries or
nodes. Each end of the duct may be connected to another duct or volume. If the end
is unconnected, the boundary is closed.
Volumes – the parts of the model where gas dynamic effects are neglected, such as
the atmosphere and cylinder. Volumes are typically larger in cross section than
ducts, and are modelled as zero dimensional thermodynamic control volumes. A
large reservoir such as the atmosphere can be modelled as an infinitely large volume
which supplies a steady pressure to any connecting ducts.
Bodies – the moving parts of the model are represented by bodies. These may be
parts such as pistons and valves. The Pempek free-piston engine had several floating
bodies which interacted through spring forces, contact and friction, such as the
mover and piston mounted passive inlet valves. Bodies can be specified in the
model to constrain the length of certain ducts, determine the volume of certain
volumes (such as the cylinders), or specify the geometric flow area through valves or
cylinder ports.
The interconnection of the various ducts and volumes must be specified, so also any
interaction between various bodies. The three main building blocks (Ducts,
Volumes and Bodies), along with the databases of interconnecting relationships are
described in more detail in Appendix VIII.
Chapter 6 The engine model – integrating all sub models 125
6.2 Calculation sequence for a complete time-step evaluation
The entire engine model operates on a common time step increment. The model
advances one time step at a time, calculating current conditions based on the
immediately preceding conditions. Since most of the pressures, flow rates, forces
and positions are interdependent, they are evaluated several times each time step in
an iterative fashion. The cylinder pressure is an especially important part of an
engine model. See Table 3-1 for the details of the volume conditions calculation.
The following list details the actual calculation sequence that was carried out for
each time step that the engine model advanced.
Shift time series data one place the right – at the beginning of a simulation, data
arrays are pre allocated with a user specified number of time series elements. Each
time the simulation advances one time step, the oldest data spills out of the array and
is lost. The current time step data occupies the lowest element (in MatLab element
1)
Calculate Bodies (1) – the first iteration simply extrapolates the velocity and
position base on the previous acceleration value. Any collisions are detected and
solved for.
Update all Duct nodal positions and velocities – this only affects ducts whose
lengths are defined by moving bodies. The initially estimated body positions are
assumed to be sufficiently accurate for these purposes.
Evaluate User Action function – this gives opportunity for special actions to be
performed. For instance the free-piston engine model models the cylinders as ducts
during the lower half of their strokes (when the gas exchange occurs). The User
Action function manages the transition from duct to volume when the mover crosses
the halfway point.
Evaluate Duct Cell properties – these are calculated from previous nodal flow
rates, flow temperature and, heat transfer.
Re-mesh the duct (if required) – test for the re-meshing criteria (see Appendix X)
Evaluate incident pressure waves – The pressure waves are advanced through
space. This calculation monitors for supersonic conditions (where a pressure wave
would be unable to propagate upstream). It also modifies the pressure wave values
for heat transfer. Mass conservation is also tested. Either cell mass is modified, or
pressure wave values are modified.
Chapter 6 The engine model – integrating all sub models 126
Calculate Volumes (1) – this first iteration of the volume calculation makes an
initial estimate of temperature by extrapolating the previous temperature rate of
change, and finding P using the ideal gas equation. The current temperature rate of
change can then be estimated based on the currently estimated volume rate of change
and previous values for flow rate and heat transfer. Current temperature and
pressure are then updated.
Evaluate flow connections (1) – with current volume pressure and temperatures
already estimated, current instantaneous flow to and from volumes is evaluated. If a
flow area coefficient must be retrieved from a map, the pressure ratio (PR) used is
averaged with the previous value, to prevent possible oscillatory situations. Flow
between ducts is also evaluated, in case the pressure here influences the motion of a
body (such a valve or piston).
Calculate Bodies (2) – this second iteration of the body motion calculation uses all
of the current values of spring force, friction, gas pressure to calculate current
instantaneous acceleration, and updates current velocity and position accordingly.
Calculate volumes (2) – volume heat transfer, chemical reactions and gas properties
are evaluated. With current flows, volume and volume rate of change known, the
current temperature rate of change is updated, allowing current temperature and
pressure to be evaluated.
Evaluate flow connections (2) – flow connections are evaluated a second time, with
updated volume conditions. Flow area coefficients are retrieved carefully using an
averaged PR with the previous iteration.
Calculate volumes (3) – the third and final volume evaluation is exactly the same as
the second iteration.
Calculate Bodies (3) – the third and final body motion calculation is exactly the
same as the second iteration. It may be unnecessary, except in cases where the body
is very sensitive to changes in volume pressure (such as a free-piston mover at peak
cylinder pressure).
Evaluate duct flow – even though the flow at internal duct nodes is not dependent
on any other parts of the model (such as volumes or bodies) this step is done last to
take advantage of having the current duct boundary flows known. This is helpful if
pressure wave information is coming from outside the duct boundary due to a
Courant number greater than 1. See Appendix IX for more details about this special
problem.
Chapter 6 The engine model – integrating all sub models 127
6.3 Programing details of the engine model
The engine model incorporating all of the sub-models was programmed using
MATLAB. A virtual engine consisted of any number of volumes, ducts and bodies,
as well as a flow connection database and a body interaction database. Every model
also had access to standard databases for gas properties, chemical reaction constants,
fuel properties, flow and force coefficients and user customisable functions. Both
the engine specific data and the standard data was implemented using the MATLAB
structure data type. This was not only convenient for readability of the code, but
also had the flexibility of allowing arrays of different size within a single data
structure. Furthermore, run-time re-sizing of data arrays was possible. This could
happen if a duct was dynamically re-meshed during simulation execution.
To assist in creating engine models, a graphical user interface was created. See
Appendix XIV for some further details of the graphical user interface.
The numeric class used for all floating point values was double precision (64bit).
Integer values were used where appropriate. Evaluation of boundary flows and
chemical equilibrium required the evaluation of a matrix division. The built-in
MARLAB function mdivide was used for this. Vectorised operations were used
where possible to speed up execution. The occurrence of runtime re-sizing or arrays
was minimised. Where possible, arguments to functions or subroutines were limited
to scalars or vectors, to save duplicating un-necessarily large data arrays in memory.
Although not implemented in the code, the structure of the model lends itself to
multi-processor execution, as many calculations can be carried out concurrently.
In total approximately 15,000 lines of code and comments were written, made up of
about 30 separate sub programs and functions. While this author chose MATLAB
due to ease of programing and de-bugging, and the availability of graphical
capability, the simulation code could otherwise have been written in any number of
programing languages such as C or Fortran.
128
. . . .
129
Chapter 7 Validations using experimental results
This chapter describes the validation work on the completed multi-purpose engine
model. First, a relatively simple model of a single shot test rig was created and
validated against an extensive set of single shot experiments conducted at Queens
university Belfast by Kirkpatrick [68]. This rig was specifically designed for
validating 1D gas dynamics codes by giving pressure data from a single generated
flow pulse without complicated wave reflections clouding the data. This has the
advantage of allowing fundamental flow processes to be easily analysed one at a
time, and decreasing the uncertainty in the experimental data about exactly what real
flow process is responsible for certain parts of the pressure record. Validation of the
model against these experiments was very valuable, as it provided a clear indication
of the real-world accuracy of the gas dynamics code in processes similar to those
found in IC engines.
Secondly, a complete model of the Pempek engine was created. It included valve
and piston dynamics, and cylinder models. The simulation results of the model are
compared to experimental engine run data; namely cylinder and compressor
pressures, and piston and exhaust valve trajectories. Clear conclusions are harder to
make here, due in part to drifting pressure signals, and also the low number of
pressure transducers on the prototype engine. The model appears to match the
limited experimental data quite well, but detailed comparison of the model results is
not possible.
Chapter 7 Validations using experimental results 130
7.1 Description of the single shot tests
7.1.1 Experimental setup of single shot rig
The single shot rig Kirkpatrick used consists of a cylinder connected to a series of
long pipes, where a single flow pulse is generated by the momentary opening of a
slide valve mechanism. The cylinder has a volume of 912cm2. The pipe is made of
aluminium and the small pipe diameter is 25mm. The slide valve has a movable
plate with an identical 25mm hole, surrounded by a sealing O-ring to ensure perfect
sealing before and after the test. The slide valve is actuated by a pneumatic impact
cylinder. In the original experiments the trajectory of the slide valve for each
actuation was carefully measured, however apart from an example case, these results
were not published. Further details of the experimental setup can be found in
Kirkpatrick’s PhD thesis [68] and in SAE papers [26, 27, 69].
The experiments were designed to test gas dynamic codes under a range of
conditions. Firstly, a code has to simulate the flow through the rapidly opening and
closing valve. The wave propagation velocity, wave distortion, attenuation and
resulting wave reflection due to friction and heat transfer was tested with straight
pipe (constant area) tests. Pressure wave transmission and reflection from passing
through a change in gas density was also tested with a special straight pipe
incorporating a second slide valve. A range of area changes were also tested, to test
for accuracy in these situations. The dimensions in mm of each of the experimental
rigs are shown in Figure 7-1 to Figure 7-8.
Figure 7-1 Straight pipe
Cylinder
slide valve
317
3691
5901
P1 P2
Chapter 7 Validations using experimental results 131
Figure 7-2 Straight pipe with density discontinuity
Figure 7-3 Sudden contraction
Figure 7-4 Convergent taper
Cylinder
slide valve
317
3097
3401
P1 P2
3703
5913
mid pipe slide valve
P3
Cylinder
slide valve
108 420
2454
P1 P2
2763 3060
P3
5270
Cylinder
slide valve
108 420
2454
P1 P2
2775 2970
P3
5481 3279
Chapter 7 Validations using experimental results 132
Figure 7-5 Sudden enlargement
Figure 7-6 Divergent taper
Figure 7-7 Short megaphone
Figure 7-8 Long megaphone
Cylinder
slide valve
317 3097
P1 P2
3394 3703
P3
6049
Cylinder
slide valve
317 3097
P1 P2
3406 3601
P3
6268 3922
Cylinder
slide valve
1705 3417
P1
3537
=84.8mm
Cylinder
slide valve
1705 3417
P1
4017
=109mm
Chapter 7 Validations using experimental results 133
7.1.2 Data compilation
Unfortunately the original data was not available, so the data was manually digitised
from plots in the original publications. The resolution and clarity of the original
graphics varied, however in most cases sufficiently good data could be recovered.
To allow easy reference to the original publications, the corresponding QUB
publication and figure number are listed in Appendix XI.
7.1.3 General model settings
The simulations were run with a global timestep of 0.1ms, and used dynamic re-
meshing. The mesh size varies and ranges from about 3.3cm up to 8cm to maintain
Courant numbers below 1.2. Where air was used in the model the composition was
21 parts Oxygen, 79 parts Nitrogen, 1 part water vapour and 0.038 parts Carbon
Dioxide. The mixture properties were calculated from the JANAF thermochemical
tables [38] as described in section 3.2. Mass conservation was imposed using the
method described in section 4.3.2. Friction was calculated as described in section
5.1.1 using a constant friction factor of 0.003. Heat transfer was calculated using the
turbulence based model described in section 5.1.4.
Chapter 7 Validations using experimental results 134
7.2 Slide valve tests
The slide valve mechanism was designed to produce a flow pulse with similar
characteristics to an engine valve or port opening event, while ensuring that perfect
sealing of the cylinder was maintained before and after the event. Unfortunately, the
trajectory of the slide mechanism has been found to be somewhat variable, and since
the trajectory data for each shot was not published, this trajectory had to be guessed
and manually adjusted for each case. This undesirable unknown casts some
uncertainty over the performance of the model. Furthermore, the experimentally
derived flow area coefficients for this valve were unclear, so a map was estimated
based on a combination of cylinder to pipe and pipe to cylinder data, and matching
with single shot results. These flow area coefficient maps are illustrated in Figure 7-
9. It would be preferable to build a flow area coefficient map directly from
experimental data, since this manually adjusted map adds a layer of uncertainty in
the fidelity of the model results.
A series of results using the straight pipe as sketched in Figure 7-1 at pressure sensor
P1 are shown in Figure 7-10 to Figure 7-19. Initial conditions in the cylinder at
release are noted in under each figure, and unless otherwise stated is air. The pipe
for all tests has a wall temperature of 293K and initially contains gas at 293K at a
pressure of 1 bar. Unless otherwise stated, the pipe initially contains air.
Most of the test cases are reproduced well by the model, though the flow area
coefficient has a strong influence on the exact shape of the pulse. The peak
pressures for the cases with different cylinder gas properties are well matched.
The characteristics of the post pulse pressure are also telling of the model’s fidelity.
Friction generates a steady reflection of the pressure wave, which for positive pulses
results in elevated post pulse pressure, as can be seen in Figure 7-10 and Figure 7-13.
Pressure waves propagating into denser gas also generate continual positive
reflections and this can be clearly seen in the air into CO2 cases (Figure 7-11 and
Figure 7-14). Conversely, heat transfer to the pipe walls always results in a
reduction in pressure wave values (both right and left travelling waves), so this effect
lowers the post pulse pressure somewhat. This balancing of friction effect and heat
transfer effect is a powerful way to calibrate models for the relative strength of
friction and heat transfer. The magnitudes of friction and heat transfer can then be
gauged by the attenuation of a pressure pulse since both effects reduce the wave
strength see section 7.4 below. Pressure waves propagating into less dense gas also
Chapter 7 Validations using experimental results 135
generate a negative reflection, as can be seen in the post pulse pressure of Figure 7-
12 and Figure 7-15.
The high pressure, high temperature release of Figure 7-17 matches the experimental
data well. The flow through the valve in this case is choked for the entire duration of
the valve opening.
The suction pulses (or rarefaction waves) shown in Figure 7-18 and Figure 7-19 also
match the experimental data well, although again, the flow area coefficient map
strongly influences the exact shape of this curve. The post pulse pressure is also
fairly well predicted.
Figure 7-20 to Figure 7-22 show the results from the so called “short pipe” shots on
the divergent taper rig which is sketched in Figure 7-4. In this rig, a short 25mm
diameter pipe is connected directly to the slide valve and then transitions to a
80.2mm diameter pipe. This situation creates a much higher pressure ratio across the
slide valve due to the presence of the nearby flow enlargement. While the suction
pulses in Figure 7-20 and Figure 7-21 are reproduced fairly well, the positive pulse
in Figure 7-22 is not. An improvement can be made by setting the flow area
coefficient to unity. It may be appropriate to modify the flow area coefficient map
so that flows at high pressure ratio are not significantly restricted. Further
investigation is necessary to identify the cause of the error. This is an important
flow type to clarify, because it is very similar to exhaust blowdown in many engines,
where an exhaust passage transitions to a larger diameter exhaust pipe in close
proximity to the valves.
Figure 7-9 Flow area coefficients used for slide valve
Cylinder to pipe flow Pipe to cylinder flow
Chapter 7 Validations using experimental results 136
Figure 7-10 Slide valve Prel =1.5 bar, Trel=293K
Figure 7-11 Slide valve Prel =1.5 bar, Trel=293K, air in cylinder, CO2 in pipe
Figure 7-12 Slide valve Prel =1.5 bar, Trel=293K, CO2 in cylinder, air in pipe
1
1.05
1.1
1.15
1.2
1.25
1.3
0 0.002 0.004 0.006 0.008 0.01
pres
sure
ratio
t (s)
model
experiment
1
1.05
1.1
1.15
1.2
1.25
1.3
0 0.002 0.004 0.006 0.008 0.01 0.012
pres
sure
ratio
t (s)
0.95
1
1.05
1.1
1.15
1.2
1.25
1.3
0 0.002 0.004 0.006 0.008 0.01 0.012
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 137
Figure 7-13 Slide valve Prel =2.4 bar, Trel=293K
Figure 7-14 Slide valve Prel =2.4 bar, Trel=293K, air in cylinder, CO2 in pipe
Figure 7-15 Slide valve Prel =2.4bar, Trel=293K, CO2 in cylinder, air in pipe
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
0 0.002 0.004 0.006 0.008 0.01
pres
sure
ratio
t (s)
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
0 0.002 0.004 0.006 0.008 0.01 0.012
pres
sure
ratio
t (s)
0.9
1
1.1
1.2
1.3
1.4
1.5
1.6
0 0.002 0.004 0.006 0.008 0.01 0.012
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 138
Figure 7-16 Slide valve Prel =2 bar, Trel=623K
Figure 7-17 Slide valve Prel =5 bar, Trel=623K
Figure 7-18 Slide valve Prel =0.5 bar, Trel=293K
0.9
1
1.1
1.2
1.3
1.4
1.5
0 0.002 0.004 0.006 0.008 0.01 0.012
pres
sure
ratio
t (s)
1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
0 0.002 0.004 0.006 0.008 0.01 0.012
pres
sure
ratio
t (s)
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
0 0.002 0.004 0.006 0.008 0.01
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 139
Figure 7-19 Slide valve Prel =0.8 bar, Trel=293K
Figure 7-20 Slide valve Prel =0.5 bar, Trel=293K short pipe shot
Figure 7-21 Slide valve Prel =0.8 bar, Trel=293K short pipe shot
0.85
0.9
0.95
1
0 0.002 0.004 0.006 0.008 0.01
pres
sure
ratio
t (s)
0.94
0.95
0.96
0.97
0.98
0.99
1
0 0.002 0.004 0.006 0.008 0.01
pres
sure
ratio
t (s)
0.965
0.97
0.975
0.98
0.985
0.99
0.995
1
0 0.002 0.004 0.006 0.008 0.01
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 140
Figure 7-22 Slide valve Prel =2.4 bar, Trel=293K short pipe shot
0.98
1
1.02
1.04
1.06
1.08
1.1
1.12
0 0.002 0.004 0.006 0.008 0.01
pres
sure
ratio
t (s)
measured
with Ca
no Ca
Chapter 7 Validations using experimental results 141
7.3 P1 driven simulation
Although the simulation results for flow through the slide valve are broadly quite
good (as shown above in section 7.2), any deviation from experiment at this point is
carried on to pressure readings downstream, and it can then be difficult to determine
where subsequent deviations from experiment originate. Furthermore, individual
shots, even at the same release pressures, sometimes displayed varying pulse profiles
due to individual variations in slide valve trajectory.
To remove this layer of uncertainty from subsequent downstream measurements, and
to expedite the task of setting up simulations (since the exact slide valve trajectory
for each shot was unknown), the experimentally recorded pressure was used to
correct the pulse profile at P1. Importantly, the complete valve and cylinder was still
functioning in the model with all of the appropriate initial properties, and release
flows. This ensured that friction, heat transfer and density reflections were properly
preserved. After the completion of the initial release pulse period, the forced
correction at P1 was turned off, and the remainder of the simulation proceeded
normally. All of the simulation results for the single shot rig results shown in the
following sections 7.4, 7.5 and 7.6 were achieved using this method.
There are a few disadvantages with this strategy however, which should be noted
here. Firstly, it depends on reliable pressure readings at P1. Quite a few pressure
readings at P1 show clear signs of ringing, and some of these are probably due to the
sensor diaphragm being unable to respond rapidly enough to sudden changes in
pressure. Secondly, some of the “bumpiness” of the pressure signal is probably due
to large scale turbulent eddies moving past the wall sensor. By clamping the
pressure at P1 to the experimentally measured values, we inadvertently introduce
these potentially spurious signals into the simulation.
Chapter 7 Validations using experimental results 142
7.4 Straight pipe shots
Figure 7-23 to Figure 7-26 show the pressure at station P2 for the straight pipe rig
shown above in Figure 7-1. Initial conditions in the cylinder at release are noted in
under each figure. The gas is air. The pipe for all tests has a wall temperature of
293K and initially contained air at 293K at a pressure of 1 bar. The pipe diameter is
25mm.
The attenuation of the pressure wave by this location in the pipe is fairly well
predicted, as is the phasing of the positive waves. The suction waves however show
a marked phase error, revealing a small inaccuracy in wave speed prediction.
Figure 7-27 shows the pressure at station P3 in the straight pipe rig shown in Figure
7-2. This test used carbon dioxide in all parts of the rig. The simulation results over
this extended period show that the wave speed in carbon dioxide is well predicted, as
well as the multiple reflections, first off the closed downstream end of the pipe, then
off the closed slide valve at the upstream end of the pipe. Wave attenuation due to
friction and heat transfer is well predicted and results in a rising pressure between
the main pulses. The slight “bumpiness” visible in these parts of the pressure record
is a numerical side-effect of the enforced mass conservation method (as described in
section 4.3.2) which is problematic for steep fronted waves when the local Courant
number is greater than 1. This problem can be reduced by decreasing the Courant
number (either reducing the time step or increasing the mesh spacing), however, for
engine modelling applications, steep fronted waves are rare.
Figure 7-28 to Figure 7-35 show the results for a pressure wave traversing a density
discontinuity in the rig sketched in Figure 7-2. The series of tests labelled AAC
have air in the cylinder and air in the first pipe section, while the second pipe section
initially contains carbon dioxide. The series labelled CCA have carbon dioxide in
the cylinder and in the first pipe section, while the second pipe section initially
contains air.
The simulation results for pressure sensor P1 correspond exactly during the initial
pulse period because they are forced to (see section 7.3 above). However after the
initial pulse passes, the simulated pressure here is released and is genuinely the result
of simulation. Crucially, the expected reflection from the downstream density
discontinuity appears with correct phasing and reasonably accurate magnitude. The
simulated results display a sharp step in the reflected pulse, due to the sharpness of
the modelled gas discontinuity. The double step is due to nearby closed cylinder
Chapter 7 Validations using experimental results 143
slide valve reflecting the “reflection” back again past pressure sensor P1. It is
probable that in the real case, some mixing of air and carbon dioxide occurred at the
mid-pipe gas density interface in the few seconds between the opening of the
separating slide valve, and the arrival of the flow pulse from the upstream cylinder.
This would explain the more diffused reflection that is apparent in the experimental
results.
The simulation results for pressure sensor P3 show that the magnitude of the
transmitted pulse is correctly modelled in the simulation.
Figure 7-23 Straight pipe P2, Prel =0.5 bar, Trel=293K
Figure 7-24 Straight pipe P2, Prel =0.8 bar, Trel=293K
Figure 7-25 Straight pipe P2, Prel =1.5 bar, Trel=293K
0.70.75
0.80.85
0.90.95
11.05
0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
0.860.88
0.90.920.940.960.98
1
0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
1
1.05
1.1
1.15
1.2
1.25
1.3
0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 144
Figure 7-26 Straight pipe P2, Prel =2.4 bar, Trel=293K
Figure 7-27 Density discontinuity P3, CCC, Prel =2.4 bar, Trel=293K, closed end
Figure 7-28 Density discontinuity P1, AAC, Prel =1.5 bar, Trel=293K
Figure 7-29 Density discontinuity P3, AAC, Prel =1.5 bar, Trel=293K
1
1.1
1.2
1.3
1.4
1.5
1.6
0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
1
1.2
1.4
1.6
1.8
0.01 0.02 0.03 0.04 0.05 0.06
pres
sure
ratio
t (s)
1
1.05
1.1
1.15
1.2
1.25
1.3
0 0.005 0.01 0.015 0.02 0.025 0.03
pres
sure
ratio
t (s)
1
1.1
1.2
1.3
0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 145
Figure 7-30 Density discontinuity P1, AAC, Prel =2.4 bar, Trel=293K
Figure 7-31 Density discontinuity P3, AAC, Prel =2.4 bar, Trel=293K
Figure 7-32 Density discontinuity P1, CCA, Prel =1.5 bar, Trel=293K
Figure 7-33 Density discontinuity P3, CCA, Prel =1.5 bar, Trel=293K
11.11.21.31.41.51.61.7
0 0.005 0.01 0.015 0.02 0.025 0.03
pres
sure
ratio
t (s)
1
1.2
1.4
1.6
0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
0.951
1.051.1
1.151.2
1.251.3
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035
pres
sure
ratio
t (s)
1
1.1
1.2
1.3
0.012 0.014 0.016 0.018 0.02 0.022 0.024
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 146
Figure 7-34 Density discontinuity P1, CCA, Prel =2.4 bar, Trel=293K
Figure 7-35 Density discontinuity P3, CCA, Prel =2.4 bar, Trel=293K
0.91
1.11.21.31.41.51.61.7
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035
pres
sure
ratio
t (s)
1
1.2
1.4
1.6
0.012 0.014 0.016 0.018 0.02 0.022 0.024
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 147
7.5 Converging flow
Figure 7-37 to Figure 7-44 show the results of a positive flow pulse traversing a
reduction in area. The sudden contraction rig is sketched in Figure 7-3, and the
convergent taper rig is sketched in Figure 7-4. The sudden contraction had a flow
area coefficient applied as shown in Figure 7-36. The small pipe diameter was
25mm and the large pipe diameter was 53mm or 80.2mm, and is listed below each
figure.
Figure 7-36 Flow area coefficients used for sudden area change
The results show that the flow behaviour for sudden contraction and gradual
contraction are very similar. The gradual taper is slightly more efficient at
transmitting a flow pulse. Simulation results match this well. Note that in the
model, converging flow is assumed to be isentropic (apart from some skin friction
and heat transfer). Also, the gradual taper has no area coefficient applied, the flow is
assumed to fill the entire cross section. Close inspection of the post pulse pressure
reveals significant ‘wigglyness’. This is caused by strong flow oscillation in the
short 108mm pipe attached immediately downstream of the slide valve, and is
reproduced quite well by the model. Note that the diverging flow model used in
these simulations was the model described in section 5.3.
From large section to small From small section to large
Chapter 7 Validations using experimental results 148
Figure 7-37 Sudden contraction 53mm P1, Prel =2.4bar, Trel=293K
Figure 7-38 Convergent taper 53mm P1, Prel =2.4bar, Trel=293K
Figure 7-39 Sudden contraction 53mm, P3, Prel =2.4bar, Trel=293K
Figure 7-40 Convergent taper 53mm, P3, Prel =2.4bar, Trel=293K
0.951
1.051.1
1.151.2
1.251.3
0 0.005 0.01 0.015 0.02 0.025
pres
sure
ratio
t (s)
0.951
1.051.1
1.151.2
1.251.3
0 0.005 0.01 0.015 0.02 0.025
pres
sure
ratio
t (s)
0.951
1.051.1
1.151.2
1.251.3
1.351.4
0.008 0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
0.951
1.051.1
1.151.2
1.251.3
1.351.4
0.008 0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 149
Figure 7-41 Sudden contraction 80.2mm, P1, Prel =2.4bar, Trel=293K
Figure 7-42 Convergent taper 80.2mm, P1, Prel =2.4bar, Trel=293K
Figure 7-43 Sudden contraction 80.2mm, P3, Prel =2.4bar, Trel=293K
Figure 7-44 Convergent taper 80.2mm, P3, Prel =2.4bar, Trel=293K
0.95
1
1.05
1.1
1.15
0 0.005 0.01 0.015 0.02 0.025
pres
sure
ratio
t (s)
0.95
1
1.05
1.1
1.15
0 0.005 0.01 0.015 0.02 0.025
pres
sure
ratio
t (s)
0.95
1
1.05
1.1
1.15
1.2
1.25
0.008 0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
0.95
1
1.05
1.1
1.15
1.2
1.25
0.008 0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 150
7.6 Diverging flow
Diverging flow presents the greatest challenge to 1D modelling, because it can
involve separated flow which is a complex phenomenon, depending on several
parameters such as geometry, flow speed and even possibly recent flow history. If
the flow is assumed isentropic, then all of the kinetic energy that is lost as a result of
the fluid slowing down is converted into increased pressure. At the other extreme,
the flow could be assumed “constant pressure” or more accurately equal pressure.
Blair [ref] recommends the use of the “momentum equation” to approximate
pressure recovery in tapered diffusing flows. A model which achieves a similar
approximation was developed and is described in section 5.3. In this thesis it is
called the pressure recovery model. In the following test cases three diffusing flow
models are tested – isentropic, pressure recovery, and equal pressure.
Figure 7-45 to Figure 7-60 show the results of a positive flow pulse traversing an
increase in area. The sudden enlargement rig is sketched in Figure 7-5 and the
divergent taper is sketched in Figure 7-6. The small pipe diameter was 25mm and the
large pipe diameter was 53mm or 80.2mm or 105.6mm, and is listed below each
figure. Note that no flow area coefficient is applied to these diverging flows.
The pressure record at sensor P1 shows the returning suction wave. In the case of
sudden enlargements, the equal pressure model is a good approximation. In the case
of gradual enlargements, the isentropic model is a good approximation. In both
cases the pressure recovery model is superior. The transmitted pressure pulse
recorded at sensor P3 shows that in most cases, the isentropic model gives the best
correspondence with the experiment. In fact all of the models under predict the peak
pressure of the transmitted pulse. A particularly puzzling result can be seen when
comparing signals at P1 and P3 for the sudden enlargement to 80.2mm at 1.5 bar
(Figure 7-51 and Figure 7-53). At P1, the isentropic model significantly over-
predicts the strength of the suction pulse, however at P3, the isentropic model
produces by far the best results. The cause of this result is unknown.
The small spike in the simulation results of the divergent taper in Figure 7-56 at
about 13.5ms is a numerical artefact of the model automatically re-meshing the
tapered portion of the duct. This issue does not affect straight sections, and could be
avoided by fixing the number of cells in tapered ducts.
Figure 7-57 and Figure 7-59 show the pressure at sensor P2 for the divergent taper.
These results are interesting because they show the reflected suction pulse partly
Chapter 7 Validations using experimental results 151
superimposed on the original pressure pulse. The pressure recovery model is clearly
superior at this position.
Figure 7-61 and Figure 7-62 show the results for a positive flow pulse traversing an
open ended divergent taper or “megaphone” – both a short, steep taper and a longer,
shallow taper. These flow rigs are sketched in Figure 7-7 and Figure 7-8
respectively. In the case of the short megaphone with an included angle of 280 the
equal pressure model give good results, while the isentropic model over-predicts the
strength of the suction pulse, suggesting that the flow is largely separated in this
case. In the case of the long megaphone with an included angle of 80 the isentropic
model gives good results suggesting that the flow remains largely attached in this
case. In both cases, the pressure recovery model also gives good results. The
secondary reflections off the closed upstream slide valve are also well predicted in
both phase and magnitude.
Figure 7-63 to Figure 7-68 show the results of a negative flow pulse traversing a
decrease in area. In this case the flow is toward the cylinder and therefore is also a
diffusing flow. In contrast to the case of positive flow pulse through an increasing
area, there is little difference between the three diffusing flow models.
Figure 7-69 and Figure 7-70 shows the sort and long megaphone tests with three
different timesteps (and mesh spacing). This is a test for mesh and time step
dependency. The model employed automatic dynamic re-meshing to maintain a
target Courant number of about 1 (maximum Courant number of 1.2). All
simulations used the pressure recovery model. The results of the simulation show
minimal mesh dependency. All but the coarsest of simulations retain good
resolution. Note that for the largest time step (0.4ms), the average cell length is
about 200mm, which is much courser than that typically used in 1D engine
modelling.
Chapter 7 Validations using experimental results 152
Figure 7-45 Sudden enlargement 53mm, P1, Prel =1.5bar, Trel=293K
Figure 7-46 Divergent taper 53mm, P1, Prel =1.5bar, Trel=293K
Figure 7-47 Sudden enlargement/ Divergent taper
53mm, P3, Prel =1.5bar, Trel=293K
0.7
0.8
0.9
1
1.1
1.2
1.3
0 0.005 0.01 0.015 0.02 0.025 0.03
pres
sure
ratio
t (s)
experiment equal P P recovery isentropic
0.7
0.8
0.9
1
1.1
1.2
1.3
0 0.005 0.01 0.015 0.02 0.025 0.03
pres
sure
ratio
t (s)
1
1.02
1.04
1.06
1.08
1.1
1.12
0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
1
1.02
1.04
1.06
1.08
1.1
1.12
0.01 0.012 0.014 0.016 0.018 0.02t (s)
Chapter 7 Validations using experimental results 153
Figure 7-48 Sudden enlargement 53mm, P1, Prel =2.4bar, Trel=293K
Figure 7-49 Divergent taper 53mm, P1, Prel =2.4bar, Trel=293K
Figure 7-50 Sudden enlargement/ Divergent taper
53mm, P3, Prel =2.4bar, Trel=293K
0.4
0.6
0.8
1
1.2
1.4
1.6
0 0.005 0.01 0.015 0.02 0.025 0.03
pres
sure
ratio
t (s)
0.4
0.6
0.8
1
1.2
1.4
1.6
0 0.005 0.01 0.015 0.02 0.025 0.03
pres
sure
ratio
t (s)
1
1.05
1.1
1.15
1.2
1.25
1.3
0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
1
1.05
1.1
1.15
1.2
1.25
1.3
0.01 0.012 0.014 0.016 0.018 0.02
t (s)
Chapter 7 Validations using experimental results 154
Figure 7-51 Sudden enlargement 80.2mm, P1, Prel =1.5bar, Trel=293K
Figure 7-52 Divergent taper 80.2mm, P1, Prel =1.5bar, Trel=293K
Figure 7-53 Sudden enlargement/ Divergent taper
80.2mm, P3, Prel =1.5bar, Trel=293K
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
0 0.005 0.01 0.015 0.02 0.025 0.03
pres
sure
ratio
t (s)
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
0 0.005 0.01 0.015 0.02 0.025 0.03
pres
sure
ratio
t (s)
0.99
1
1.01
1.02
1.03
1.04
1.05
1.06
0.01 0.012 0.014 0.016 0.018
pres
sure
ratio
t (s)
0.99
1
1.01
1.02
1.03
1.04
1.05
1.06
0.01 0.012 0.014 0.016 0.018t (s)
Chapter 7 Validations using experimental results 155
Figure 7-54 Sudden enlargement 80.2mm, P1, Prel =2.4bar, Trel=293K
Figure 7-55 Divergent taper 80.2mm, P1, Prel =2.4bar, Trel=293K
Figure 7-56 Sudden enlargement/ Divergent taper
80.2mm, P3, Prel =2.4bar, Trel=293K
0.4
0.6
0.8
1
1.2
1.4
1.6
0 0.005 0.01 0.015 0.02 0.025 0.03
pres
sure
ratio
t (s)
0.4
0.6
0.8
1
1.2
1.4
1.6
0 0.005 0.01 0.015 0.02 0.025 0.03
pres
sure
ratio
t (s)
0.98
1
1.02
1.04
1.06
1.08
1.1
1.12
1.14
1.16
0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
0.98
1
1.02
1.04
1.06
1.08
1.1
1.12
1.14
1.16
0.01 0.012 0.014 0.016 0.018 0.02
t (s)
Chapter 7 Validations using experimental results 156
Figure 7-57 Divergent taper 80.2mm, P2, Prel =2.4bar, Trel=293K
Figure 7-58 Divergent taper 105.6mm, P1, Prel =2.4bar, Trel=293K
Figure 7-59 Divergent taper 105.6mm, P2, Prel =2.4bar, Trel=293K
0.6
0.8
1
1.2
1.4
1.6
0.007 0.009 0.011 0.013 0.015 0.017 0.019
pres
sure
ratio
t (s)
0.4
0.6
0.8
1
1.2
1.4
1.6
0 0.005 0.01 0.015 0.02 0.025 0.03
pres
sure
ratio
t (s)
0.6
0.8
1
1.2
1.4
1.6
0.007 0.009 0.011 0.013 0.015 0.017 0.019
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 157
Figure 7-60 Divergent taper 105.6mm, P3, Prel =2.4bar, Trel=293K
Figure 7-61 Short Megaphone P1, Prel =2.0bar, Trel=293K
Figure 7-62 Long Megaphone P1, Prel =2.0bar, Trel=293K
1
1.02
1.04
1.06
1.08
1.1
1.12
0.01 0.012 0.014 0.016 0.018 0.02
pres
sure
ratio
t (s)
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
pres
sure
ratio
t (s)
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
pres
sure
ratio
t (s)
Chapter 7 Validations using experimental results 158
Figure 7-63 Sudden contraction 53mm, P1, Prel =0.5bar, Trel=293K
Figure 7-64 Convergent taper 53mm, P1, Prel =0.5bar, Trel=293K
Figure 7-65 Sudden contraction / Convergent taper
53mm, P3, Prel =0.5bar, Trel=293K
0.86
0.88
0.9
0.92
0.94
0.96
0.98
1
0 0.005 0.01 0.015 0.02 0.025
pres
sure
ratio
t (s)
0.86
0.88
0.9
0.92
0.94
0.96
0.98
1
0 0.005 0.01 0.015 0.02 0.025
pres
sure
ratio
t (s)
0.75
0.8
0.85
0.9
0.95
1
0.008 0.01 0.012 0.014 0.016 0.018
pres
sure
ratio
t (s)
0.75
0.8
0.85
0.9
0.95
1
0.008 0.01 0.012 0.014 0.016 0.018t (s)
Chapter 7 Validations using experimental results 159
Figure 7-66 Sudden contraction 53mm, P1, Prel =0.8bar, Trel=293K
Figure 7-67 Convergent taper 53mm, P1, Prel =0.8bar, Trel=293K
Figure 7-68 Sudden contraction / Convergent taper
53mm, P3, Prel =0.8bar, Trel=293K
0.93
0.94
0.95
0.96
0.97
0.98
0.99
1
0 0.005 0.01 0.015 0.02 0.025
pres
sure
ratio
t (s)
0.93
0.94
0.95
0.96
0.97
0.98
0.99
1
0 0.005 0.01 0.015 0.02 0.025
pres
sure
ratio
t (s)
0.88
0.9
0.92
0.94
0.96
0.98
1
0.008 0.01 0.012 0.014 0.016 0.018
pres
sure
ratio
t (s)
0.88
0.9
0.92
0.94
0.96
0.98
1
0.008 0.01 0.012 0.014 0.016 0.018t (s)
Chapter 7 Validations using experimental results 160
Figure 7-69 Short Megaphone P1, Prel =2.0bar, Trel=293K,
various simulation timesteps
Figure 7-70 Long Megaphone P1, Prel =2.0bar, Trel=293K,
various simulation timesteps
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
0.015 0.02 0.025 0.03 0.035 0.04
pres
sure
ratio
t (s)
measured
0.0004
0.0002
0.0001
0.00005
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2
0.015 0.02 0.025 0.03 0.035 0.04
pres
sure
ratio
t (s)
measured
0.0004
0.0002
0.0001
0.00005
Chapter 7 Validations using experimental results 161
7.7 Modelling the Pempek engine
The complete engine model as described in chapters 3, 4,5 and 6 was developed with
its primary purpose being to model the Pempek free-piston engine. In this section,
the Pempek engine is modelled, with a view to check the model’s fidelity against
experimentally measured data (by Pempek). If the model is able to reproduce the
measured engine behaviour reliably, then it can also be used with reasonable
confidence to do predictive simulations.
One further benefit of such a model is that it can be used in conjunction with
physical engine testing to illuminate the detailed inner workings of the engine, in
particular gas dynamics and in the case of the Pempek engine valve motion.
Therefore this section closes with further model results which are instructive, even
though they aren’t validated against direct measurements.
7.7.1 Engine model setup
Figure 7-71shows the complex 3D shape of the inlet side ducting of the Pempek
engine. In actual fact, the compressor volume is even more complicated than the
simple prism shown. Also, the four exhaust valve passages (not shown) turn through
90 degrees and are junctioned together immediately downstream.
Figure 7-71 Pempek engine inlet side ducting half section 3D view
To create the 1D engine model, each part of the three dimensional engine ducting is
approximated as a straight duct section. Figure 7-72 shows the layout of the 1D
Pempek engine model. The ducting for only one cylinder is shown, since the inlet
and exhaust systems for each cylinder are completely independent. The leftmost
duct section is the atmospheric intake pipe which connects to the internal compressor
compressor volume
4 x passive inlet valves
4 x exhaust valves
Combustion cylinder
Chapter 7 Validations using experimental results 162
volume via a check valve. The rightmost duct is the end of the exhaust pipe and
terminates at the atmosphere. The length of each duct section is shown in mm above
the graphic. The length of the compressor volume and combustion cylinder is
dynamic and is determined by the instantaneous position of the mover. The cross
sectional area of each duct section is shown in mm2 below the graphic. Each duct
section is constant area.
Figure 7-72 Pempek engine1D model layout
The atmospheric temperature and pressure was 300K and 101325Pa. The
atmospheric composition was 21 parts Oxygen, 79 parts Nitrogen, 1 part water
vapour and 0.038 parts Carbon Dioxide. The chemical composition of the (diesel)
fuel was C10.8H18.7. Duct friction and heat transfer were modelled using the methods
described in section 5.1. The combustion cylinder was modelled as a duct during the
lower half of the piston motion. It was modelled as a thermodynamic control
volume during the high pressure part of the cycle. Cylinder heat transfer, blowby,
fuel injection and combustion were modelled using the methods described in section
5.2. The motion of the mover, inlet valves and exhaust valves were determined by
the equation of motion (5-28) and the multi-body dynamics model described in
section 5.5. The motion of the exhaust valve armature was prescribed based on
experimental data. The mass conservation option for 1D gas dynamics was turned
on. The model time step was 0.1ms. Dynamic re-meshing was employed.
The engine model was set up with the same control signals as the real engine. Data
from a single engine run was used for comparison against the model. The engine run
began with four un-fired cycles, followed by 14 fired cycles. The fourth unfired
100 180 190 80 160 105 60 60 42 100 76±x 56±x
combustion cylinder exhaust pipe
compressor volume
600
mm
mm2 5238 1000 728 1900
3421 1772 804 1500 1777 1521 1521
compressor check valve
passive inlet valves
exhaust valves
Chapter 7 Validations using experimental results 163
cycle and the fourth fired cycle are compared to the model, which was cycled several
times to achieve approximately steady operation.
7.7.2 Results compared to experiment
Results for the model compared to experiment for both motored and fired cycles are
shown in Figure 7-73 to Figure 7-78. Only the right cylinder is shown.
The experimental pressure measurements from the cylinder and compressor have an
unknown DC offset, so the absolute pressure must be guessed. Furthermore, some
short term drift in the pressure signal may be present. Mover position was inferred
from instantaneous generator coil voltages. See the footnotes on page 21 for more
information on the cylinder pressure and mover position instrumentation.
Figure 7-73 Motored engine comparison with experiment
-60
-40
-20
0
20
40
60
0
1
2
3
4
5
6
-0.005 0 0.005 0.01 0.015 0.02 0.025 0.03
Posi
tion
(mm
)
P (b
ar)
t (s)
mover position
compressor pressure
cylinder pressure
model
experiment
Chapter 7 Validations using experimental results 164
Figure 7-74 Motored engine, indicator diagram
Figure 7-75 Motored engine, indicator diagram
0
10
20
30
40
50
60
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
P (b
ar)
position (mm)
model
experiment
0
1
2
3
4
5
6
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
P (b
ar)
position (mm)
model
experiment
Chapter 7 Validations using experimental results 165
Figure 7-76 Fired engine comparison with experiment
Figure 7-77 Fired engine, indicator diagram
-60
-40
-20
0
20
40
60
0
1
2
3
4
5
6
-0.005 0 0.005 0.01 0.015 0.02 0.025
Posi
tion
(mm
)
P (b
ar)
t (s)
mover position
cylinder pressure
compressor pressure
model
experiment
0
10
20
30
40
50
60
70
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
P (b
ar)
position (mm)
model
experiment
Chapter 7 Validations using experimental results 166
Figure 7-78 Fired engine, indicator diagram
The model mover trajectory is not a perfect match with the experiment. This is
probably due to velocity signal smoothing in the real engine control system causing a
somewhat different generator force profile. Notwithstanding the minor difference in
trajectory, the model predicts the stroke length and frequency quite well for both
motored and fired cases.
The pressure at the compressor volume is a useful point of comparison. It indirectly
reflects the flow of inlet air into the combustion cylinder through the passive inlet
valves. The position of the pressure sensor is at the extremity of the compressor
volume, so changes in pressure here lag behind changes in pressure near the inlet
valves. The model reproduces the compressor pressure quite well, considering that it
is modelled with a simplified arrangement of ducts. This good match with
experimental data suggests that the passive inlet valve flows are broadly accurate.
The cylinder pressure during the gas exchange period also matches the experimental
data reasonably well. This is further evidence that the modelled trajectory of the
passive inlet valves is broadly accurate. The cylinder pressure in the final stages of
the gas exchange period (just before the exhaust valves close is less well predicted,
suggesting that the exhaust gas dynamics contain some inaccuracies. This is not
surprising, since the gas dynamics in the exhaust pipe are quite complex. In the case
of the fired engine, exhaust blowdown generates a flow pulse, which when reflected
off the open end of the pipe, returns as a suction pulse to the exhaust valves. This
0
1
2
3
4
5
6
7
8
9
10
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
P (b
ar)
position (mm)
model
experiment
Chapter 7 Validations using experimental results 167
suction pulse reflects off the semi restricted valve area, and returns again to the open
pipe end. On reaching the pipe end, a weak pressure pulse returns to the cylinder,
reaching it just before the exhaust valves close, forcing a little gas back into the
cylinder. The final cylinder pressure is affected by the precise timing of the exhaust
pressure waves, the exhaust valve closing, as well as the way the inlet valve flow
contributes to the exhaust flow. Animation movies of the motored and fired
simulations are available on disk and are listed in Appendix XV.
Close examination of the indicator diagram for the motored case (Figure 7-74 and
Figure 7-75) shows that there is substantial loss of cylinder pressure during the
closed part of the cycle, resulting in a large area of negative work. Though part of
this effect is due to heat transfer, the largest contributor in this case is likely cylinder
leakage (blowby). Note that sensor drift may also contribute to this result. (See
footnote on page 21) Assuming the measurements are reasonably accurate, however,
the blowby rate estimated in the model appears to be too small.
7.7.3 More model results
The model gives access to details about engine operation that are not otherwise
available. This section gives a selection of modelled engine data to illustrate the
useful role of the engine model to interpret otherwise scanty experimental
measurements.
Figure 7-79 shows the simulated exhaust and inlet valve trajectories. It is interesting
to note the trajectory of the piston mounded passive inlet valves, as they respond to
the combined influence of gas pressure and a rapidly accelerating piston near BCD.
The figure also shows the mass flows through inlet and exhaust valves. It is
interesting to note the momentary reverse flow at the exhaust valve at around 19ms.
Chapter 7 Validations using experimental results 168
Figure 7-79 Valve trajectories and mass flows
Figure 7-80 shows the simulated mass of gas in the cylinder, and the cumulative total
of the supplied inlet air. According to the simulation, the delivery ratio is about 1.6,
which should produce good scavenging efficiency. It will also mean that a large
amount of cool, oxygen rich air enters the exhaust pipe after each cycle. The
simulated blowby results in a substantial loss of cylinder mass during the
compression stroke, and the early stages of combustion.
Figure 7-80 Cylinder mass and inlet mass
Figure 7-81 shows the simulated cylinder chemical composition. Note that nitrogen
is not shown on this plot. Fuel injection begins at about -1.5ms and combustion
begins at about 0.3ms. As combustion proceeds, the oxygen and fuel fractions fall,
while the carbon dioxide and water fractions increase. The mass of fuel that was
-0.025
0
0.025
0.05
0.075
0.1
0.125
0.15
0.175
0.2
0.225
0.25
0.275
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0 0.005 0.01 0.015 0.02 0.025
mas
s flo
w ra
te (g
/ms)
posi
tion
(m)
t (s)
Exhaust valve armature (x10)
mover, inlet valve position
exhaust valve (x10) inlet valve
(x10)
exhaust mass flow rate
inlet mass flow rate
-0.06
0
0.06
0E+0
1E-4
2E-4
3E-4
4E-4
5E-4
-0.005 0 0.005 0.01 0.015 0.02 0.025
posi
tion
(m)
mas
s (kg
)
t (s)
cylinder mass
cumulative inlet mass
mover, inlet valve position
Chapter 7 Validations using experimental results 169
injected (15mg) results in a slightly rich mixture. Thus late in the combustion
process, once oxygen has been consumed, a small amount of carbon monoxide and
hydrogen is produced. Finally, during the gas exchange period, the spent
combustion gases are purged from the cylinder and replaced with a fresh air charge.
The model has about 8% residual exhaust remaining, though this result depends on
the mixing coefficient used for the cylinder flow, and can be adjusted by the user
(see section 4.5.2).
Figure 7-82 shows the simulated cylinder temperature and the gas mixture ratio of
frozen specific heats ( ). A large drop in happens with the introduction of the
fuel. The value of this gas property for air at room temperature is 1.4, but is
substantially lower for gas mixtures inside an engine under operating conditions.
See Appendix II for a short discussion on frozen specific heats.
Figure 7-81 Cylinder species mass fractions
Figure 7-82 Cylinder temperature and specific heat ratio
0%
5%
10%
15%
20%
-0.005 0 0.005 0.01 0.015 0.02
mas
s fra
ctio
n
t (s)
CO2 O2
H2O
CO
H2 fuel
1.2
1.25
1.3
1.35
1.4
0
250
500
750
1000
1250
1500
1750
2000
2250
-0.005 0 0.005 0.01 0.015 0.02
K
t (s)
cylinder temp
ratio of specific heats (frozen)
170
. . . .
171
Chapter 8 Predictive modelling
This chapter applies the multi-purpose engine model to two proposed design
modifications of the Pempek engine.
The existing engine had a high pressure compressor to ensure that sufficient air is
delivered to the cylinder. However, analysis of the compressor pressure shows that
it consumes almost 30% of the engine’s indicated work (section 2.4). Part 1 of this
chapter applies the gas dynamics engine model to the existing engine in order to
determine whether it is possible to lower the compressor pressure, but maintain
sufficient air delivery through the original passive inlet valves.
Part 2 of this chapter describes a possible radical re-design of the engine - back to a
more typical port inlet layout. Without the integrated compressor, alternative means
of driving the scavenging process must be provided. Part 3 examines the feasibility
of using exhaust pressure wave energy to drive the scavenging.
Chapter 8 Predictive modelling 172
8.1 Optimising the existing layout
A slightly modified version of the existing engine was designed and then tested
using the model. The clearance volume of the compressor was increased to lower
the compressor pressure (this could be achieved by installing pistons with more
internal cavity volume). To help the passive inlet valves open, the return spring
strength was reduced from 6.8N/mm to 0.5N/mm, and the closed position spring
compression increased from 0.1mm to 1mm. The exhaust valve opening period was
reduced from 10 ms to 7 ms. For simplicity, exhaust valve timing and injection
timing were fixed for all load levels. A tuned exhaust was added, which serves to
provide some suction effect during the scavenging period, and at high load, to
produce a plugging pulse after inlet valve closure, but just before exhaust valve
closure. The engine layout is shown in Figure 8-1.
Figure 8-1 Tuned exhaust pipe layout for existing engine
All other model settings are the same as described above in section 7.7.1. The
engine model was tested across the load range from 0-22 mg of injected fuel per
cycle and the results are given in Figure 8-2. As a result of the earlier closing
exhaust valve, the somewhat cooler inlet air (because of a lower compressor work),
and a slightly elevated final cylinder pressure, the mass of air trapped in the cylinder
was increased from around 280mg to 400mg. This allowed the maximum fuel
quantity to be increased from 14.5mg to 22mg. Since the modelled blowby was the
same as used for the existing engine, higher combustion pressure resulted in higher
blowby loss, which reduces indicated efficiency at higher load. The indicated work
for the modified compressor was 22 joules (compared to 48 joules for the existing
engine). The indicated efficiency is based on a fuel heating value of 43MJ/kg, and
the net indicated efficiency has the compressor work subtracted from the cylinder
work.
300 300 230 300 300 550
combustion cylinder
exhaust pipe
compressor volume
Chapter 8 Predictive modelling 173
Figure 8-2 Results for optimised layout
Figure 8-3 shows the valve trajectories and mass flows for the engine at maximum
fuelling. The exhaust valves open slightly earlier than the existing engine (Figure 7-
79 above), however due to the lower compressor pressure, the passive inlet valves
open a little later than in the existing engine, and only have one main opening.
Figure 8-3 Valve trajectories and mass flows (full power)
Figure 8-4 shows the mass of gas in the cylinder, and the cumulative total of the
supplied inlet air. According to the simulation, the delivery ratio is about 1.2, which
should ensure reasonably effective scavenging. It also means that some cool,
oxygen rich gas mixture enters the exhaust pipe after each cycle. The simulated
blowby results in a substantial loss of cylinder mass during the compression stroke,
and the early stages of combustion.
0
100
200
300
400
500
600
0 5 10 15 20
mas
s (m
g)
mg fuel
air mass delivered
10%
20%
30%
40%
50%
0 5 10 15 20-50
050
100150200250300350
mg fuel
ener
gy (J
)
ind. energyind. efficiencynet ind. efficiency
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25
-0.07
-0.05
-0.03
-0.01
0.01
0.03
0.05
0.07
0.01 0.015 0.02 0.025 0.03
mas
s flo
w ra
te (g
/ms)
Posi
tion
(m)
t (s)
exhaust valve armature (x10)
mover, inlet valve position
exhaust valve (x10) inlet valve (x10)
exhaust mass flow rate
inlet mass flow rate
Chapter 8 Predictive modelling 174
Figure 8-4 Cylinder mass and inlet mass (full power)
Figure 8-5 shows the cylinder chemical composition. Note that for clarity nitrogen
is not shown on this plot. Fuel injection begins at about 8.4 ms and combustion
begins at about 10.6 ms.
Figure 8-5 Cylinder species mass fractions (full power)
There is sufficient oxygen in the cylinder to burn 22mg of the diesel fuel. However
in actual fact, the maximum fuel rate would probably be less than this due to the
finite mixing rate of fuel spray. The maximum gas temperature with a full 22 mg of
fuel is about 2160 K, which may also produce unacceptably high NOx emissions.
Thus, on this measure too, maximum fuelling may have to be reduced somewhat.
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0E+0
1E-4
2E-4
3E-4
4E-4
5E-4
6E-4
0.005 0.01 0.015 0.02 0.025 0.03
posi
tion
(mm
)
mas
s (kg
)
t (s)
cylinder mass
cumulative inlet mass
mover, inlet valve position
0%
5%
10%
15%
20%
0.005 0.01 0.015 0.02 0.025 0.03
mas
s fra
ctio
n
t (s)
CO2 O2
H2O fuel
Chapter 8 Predictive modelling 175
8.2 Port admission layout
The Pempek engine has an integral compressor and utilises passive inlet valves
mounted on the head of the piston. This novel mechanism takes away the need for
cylinder ports which can accelerate piston ring wear and increase oil consumption.
The overall engine package is extremely compact owing to the overlapped generator.
However, the passive inlet valves present some inherent difficulties. The trajectory
of the valves is not fixed and depends on both piston motion and pressure differential
across the valve. (both exhaust valve timing and compressor pressure). Also, long
term durability has not been demonstrated, and this is all the more questionable
given the repeated seating impacts, heat, and unlubricated location. Furthermore, the
passive inlet valves constrain the shape of the piston head compared to what would
otherwise be possible. The flow of inlet air through internal cavities of the piston
will result in heat transfer to the inlet charge which will reduce charging efficiency
and may increase NOx emissions.
A radical re-design of the Pempek engine using inlet ports is described in this
section. Figure 8-6 shows a cross section of the existing Pempek engine alongside a
modified design with cylinder inlet ports. Passive inlet valves, internal mover
cavities and compressor check valves are eliminated.
Figure 8-6 Modification of Pempek engine for port admission
Existing Pempek Engine
Modified engine with port admission
Chapter 8 Predictive modelling 176
As with the original design, the mover is supported by the generator stator and the
piston does not directly touch the cylinder. Piston rings seal the small gap between
piston and cylinder and need some form of lubrication.
Of course, such a design modification takes away much of what makes the Pempek
engine unique (integral compressor and no cylinder ports). The issue of piston ring
lubrication and wear returns, and some form of external supercharging may be
necessary.
8.2.1 Contactless piston
The current Pempek engine experimented with an alternative cylinder sealing
method (as explained above in section 5.2.2) that used a cylinder ring at the bottom
of the cylinder liner. Although this resulted in unacceptably high cylinder leakage,
the blowby was not excessively large, and this result begs the question of whether
further refinement of a contactless ‘seal’ might not be possible.
A contactless piston would have to have a very precise fit to the cylinder, and be
accurately supported and aligned. The main mover bearings of the existing engine
consist of engineered plastic pads running on the large metal stator surface. (This
method was possible given the very low bearing pressure) However, a small bearing
clearance is necessary to ensure that the mover does not bind and to allow for some
thermal expansion of the mover. This clearance reduces the accuracy with which the
pistons can be aligned within the cylinder. Manufacturing tolerances of the complex
mover assembly, generator stator and cylinder liner further reduce piston alignment
accuracy.
One bearing technology that is capable of precise positioning is air bearings. For
example, the port admission layout shown above in Figure 8-6 could have an air
bearing pad at the bottom of the cylinder, with the piston being running surface. An
alternative design is shown in Figure 8-7 where the rear of the magnet holder
becomes the bearing surface. In both of these designs, and the cylinder structure
supports the mover. This frees the air gap part of the generator from being the
bearing surface, which is beneficial since neither the slotted stator surface nor the
surface mounted magnets on the mover are ideal bearing surfaces. Compressed air is
supplied to the air bearing pads.
Chapter 8 Predictive modelling 177
Figure 8-7 Contactless pistons with air bearings
There is much that is doubtful about the designs suggested above. For instance, can
the pistons/mover be adequately cooled? Can the piston-cylinder gap be controlled
sufficiently (given differential thermal expansion) to maintain engine efficiency but
prevent catastrophic seizing? How closely, for that matter, can the air bearing
clearance be maintained? How strong might un-balanced pressure forced on the
piston sides be? How much parasitic power would be required to supply the air
bearings with compressed air?
Leaving aside for the time being the feasibility of such a contactless piston engine, it
is clear that such an engine could be almost maintenance free, and there would be no
need for lubricating oil.
8.2.2 External supercharging
The loss of the integral compressor means that some other method must be found to
charge the cylinder with fresh air every cycle. An electric supercharger to provide
low-pressure boost would be suitable. The constant speed nature of a free-piston
engine suggests that gas dynamic tuning could also be utilised to provide beneficial
air flow to the engine. Hibi and Ito [63] have proposed removing the exhaust
ventilation fan (that drove the cylinder charging process) and using in its place the
inertia effect of exhaust gas. The rest of this chapter applies the gas dynamics
engine model to a port admission version of the Pempek engine, in order to
investigate the engine performance over the full load range and to assess the viability
of using exhaust system gas dynamics to drive the scavenging process.
mover exhaust valves
inlet bearing surface
fresh air in
stator
Chapter 8 Predictive modelling 178
8.3 Gas dynamics driven scavenging
8.3.1 Model setup
The same mover mass and piston bore and stroke of the Pempek engine were used
for the new port scavenged layout. Inlet port opening was determined by mover
position, and so was by definition symmetric around BDC. Many prototype free-
piston engines use exhaust ports too. Most are loop scavenged [7, 15, 75, 76, 114,
115, 125], though a couple are opposed piston uni-flow scavenged [63, 120]. In the
present model, a symmetrical, mover controlled exhaust opening was used so that
results here could also be applied to these other engines. Of course, more timing
flexibility would be available with poppet valve exhaust.
To simulate the flow restrictions due to an air filter, the intake ports were connected
to a large plenum (20 litres) with a restricted inlet from the atmosphere. To simulate
the flow exhaust flow restrictions such as a muffler and catalytic converter, the
exhaust system was connected to a large plenum (6 litres) with a restricted flow to
the atmosphere. The intake and exhaust plenum pressures are given in the modelling
results below, and represent the adverse pressure differential which must be
overcome to charge the engine by gas dynamics alone. Figure 8-8 shows the layout
of the port admission engine model. The length of each duct section is indicated
above the graphic, and the cross sectional area at each section is indicated below the
graphic. A short inlet passage leads from the inlet plenum (not shown) to the
cylinder. The exhaust pipe is made from a series of tapered ducts and terminates at
the exhaust plenum (not shown). Only one cylinder is shown, but the engine model
has two independent cylinders and exhaust systems.
Figure 8-8 Port admission engine model layout
The cylinder blowby was modelled at ¼ of the rate used to model the existing
engine. The exhaust pipe wall temperature was set at a constant 500 K. All other
model settings are the same as described above in section 7.7.1
850 300 400 260 480 400
cylinder exhaust pipe
200
1667 mm2 3421
1200 1100 1500 3000 6362 400 6362
mm
Chapter 8 Predictive modelling 179
8.3.2 Results
The engine model was run with a range of fuelling levels, from 2.5mg of injected
fuel , up to 22.5 mg. The model typically took at least 4 cycles to settle to a steady
operating state after a change in fuel level. Figure 8-9 shows the general results for
the port admission layout. The engine relies on exhaust pressure each cycle to re-
charge the cylinder, so 2.5mg of fuel is barely enough to sustain sufficient air
delivery. Furthermore, pressure loss due to blowby and heat transfer becomes a
significant proportion at low fuel levels. However at 5mg of fuel and higher, the
engine is easily able to charge. Operating frequency increases with fuel. The slow
decline in indicated efficiency at higher fuelling levels is mainly attributable to the
increased specific heat of high temperature, combustion product rich mixture, though
increased ‘time loss’, blowby and heat transfer may also contribute.
Figure 8-9 General results for port admission layout
Figure 8-10 shows the port openings and mass flows for the engine at full power
setting. Figure 8-11 shows the cylinder pressure for the same cycle. The initial
rapid blowdown of the cylinder draws the cylinder pressure slightly below the intake
pressure. A suction condition at the exhaust port is maintained for about 3.5 ms after
which the inlet port closes. Soon after this, a so called ‘plugging pulse’ arrives at the
exhaust port and rapidly forces some of the recently exhausted gas back into the
cylinder, boosting the pressure to around 1 bar above atmospheric. Figure 8-12
shows modelled the inlet plenum and exhaust plenum pressures over one cycle.
Note that the plenums are connected to both cylinders.
Figure 8-13 shows the cylinder mass, oxygen mass and cumulative inlet flow.
0
100
200
300
400
500
600
0 5 10 15 20
mg
mg fuel
masstrapped
air massdelivered
0 5 10 15 200
5
10
15
20
25
30
35
40
mg fuel
Hz frequency
20%
30%
40%
50%
60%
0 5 10 15 200
50
100
150
200
250
300
350
400
mg fuel
J
ind. energy
ind. efficiency
Chapter 8 Predictive modelling 180
Figure 8-10 Port openings and mass flows (full power)
Figure 8-11 Cylinder pressure during scavenging (full power)
Figure 8-12 Modelled inlet and exhaust plenum pressure (full power)
-0.07
-0.05
-0.03
-0.01
0.01
0.03
0.05
0.07
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25
0.01 0.0125 0.015 0.0175 0.02 0.0225 0.025 0.0275
Posi
tion
(m)
mas
s flo
w ra
te (g
/ms)
t (s)
exhaust port opening
mover position inlet port
opening
exhaust mass flow rate
inlet mass flow rate
0.5
1
1.5
2
2.5
3
0.015 0.0175 0.02 0.0225 0.025
P (b
ar)
t (s)
cylinderinlet plenumexhaust plenum
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
0 0.005 0.01 0.015 0.02 0.025
P (b
ar)
t (s)
atmosphere
inlet plenum
exhaust plenum
Chapter 8 Predictive modelling 181
Figure 8-13 Cylinder mass and inlet mass (full power)
Figure 8-14 shows a comparison of cycle temperatures for the range of fuelling
levels. Interestingly, the lowest fuelling level (2.5mg) actually had higher cylinder
temperatures due to low air delivery mass and high levels of residual exhaust gas.
Figure 8-14 Cycle temperatures
0E+00
1E-04
2E-04
3E-04
4E-04
5E-04
6E-04
0.015 0.0175 0.02 0.0225 0.025
mas
s (kg
)
t (s)
cylinder mass
cumulative inlet mass
O2 mass
250
500
750
1000
1250
1500
1750
2000
2250
0 0.02 0.04 0.06 0.08 0.1
T (K)
position (m)
22.5mg
20mg
15mg
175.mg
12.5mg
10mg
7.5mg 5mg
2.5mg
Chapter 8 Predictive modelling 182
Figure 8-15 shows the cycle pressures for various fuelling levels. The lowest
fuelling level suffers from low cylinder mass due to the absence of any plugging
pulse. Figure 8-16 shows a comparison of cylinder pressure during scavenging. The
clear trend is for increasing strength of plugging pulse with increasing combustion
energy.
Figure 8-15 Cycle pressures
Figure 8-16 Cycle pressures during scavenging
0
20
40
60
80
100
0 0.02 0.04 0.06 0.08 0.1
bar
position (m)
22.5mg
15mg
10mg
5mg
2.5mg
0.5
1
1.5
2
2.5
3
3.5
4
0.06 0.07 0.08 0.09 0.1 0.11
bar
position (m)
22.5mg
15mg
10mg
5mg
2.5mg
22.5mg
5mg
Chapter 8 Predictive modelling 183
8.4 Conclusion
One of the Pempek engine’s distinguishing features is the passive inlet valves and
integrated compressor, which remove the need for problematic cylinder ports.
However, the current engine uses an enormous proportion of power to compress the
inlet air to over 2 bar above ambient. The results of the simulations in section 8.1
show that it quite possible to run the engine at lower compressor pressure. This re-
design reduces the compressor work from 48J to 22J, which is equivalent to a
pumping IMEP of 0.6bar. Furthermore, improvements in trapping efficiency
increase the maximum fuelling level from about 14mg to 22mg – translating to a
power increase of over 50%. Results of modelling an un-boosted engine (section
8.3) suggests that even lower compressor pressures are entirely possible, providing
that the passive inlet valves can be designed to operate without pressure across them.
However passive inlet valves, internal mover cavities and integral compressor are
somewhat complex and constrain the design of the mover and piston crown, heat
transfer to the inlet charge will be increased, and moreover, passive inlet valves are
un-proven for long term durability. The alternative of port inlet (and often exhaust)
is used by many free-piston engine prototypes, along with traditional piston rings.
One elegant possible alternative to piston rings is a completely contactless ‘seal’,
though the feasibility of this option is unknown.
Regardless of whether a contactless piston seal can be implemented, or if traditional
lubricated piston rings are used - the simulation results in section 8.3 show that a
tuned exhaust pipe can charge a two stroke engine without any added assistance
from external blowers. Moreover, this has been demonstrated for the case of port
inlet and port exhaust which is the most restrictive case.
Simulated free-piston engine operation was fairly robust, and was able to self-charge
from full power down to about 10%. The engine cannot self-charge when motoring,
so some method of purging the cylinders with air would be needed to start the
engine. One problem encountered was that a sudden increase in fuel load sometimes
triggered un-symmetric operation, where one cylinder was unable to fully charge.
Since stronger combustion in one cylinder resulted in higher trapped mass, this
produced a corresponding reduction in stroke length to TDC, which reduced the port
opening for the other cylinder (which is at BDC). This problem could be avoided by
increasing fuel level over two cycles instead of one. Alternatively, a less aggressive
Chapter 8 Predictive modelling 184
plugging pulse may result in a more consistent trapped cylinder mass over the load
range (as opposed to the increasing trapped mass shown in Figure 8-9).
In the simulation, the entire exhaust pipe wall temperature was arbitrarily set at
500K. This is not realistic, of course, since the part of the pipe closest to the
cylinder will be significantly hotter than the downstream sections, and this will
depend on a range of factors like exhaust gas temperature and whether the pipe is
externally insulated. Since pressure wave speed is proportional to the square root of
gas temperature, modelling the complex interactions of continual wave pulsations in
the pipe depends on accurate gas temperature modelling. It may require some
experimental iterations with prototype pipes to refine the model for wall temperature
etc.
The tuned exhaust pipe which assists scavenging (in both cases analysed) is over 2m
long and may be difficult to package with the engine, though the pipe can be bent
around on itself without significant loss of performance (as is done, for example with
2-stroke motorcycles exhausts). The pipe could be made somewhat shorter (say 1m
shorter) if the strong plugging pulse was not needed. A weak plugging pulse would
still be available, but maximum engine power would be somewhat less (say 20%).
Alternatively, if the exhaust was controlled by poppet valves, then the exhaust valves
could close momentarily before the inlet port closed, preventing excessive loss of
cylinder mass to the exhaust. It may also be possibility is to have both cylinders
supplying a single tuned pipe exhaust.
The simulations that were done above paid no attention to intake system gas
dynamics, but it is likely that additional gains in charging efficiency can be made
here.
All this is to say, in the end, that a myriad of design options are available to take
advantage of a free-piston engine’s inherent constant speed nature, for improving or
controlling the charging of the engine. The complexity of wave interactions in
engine ducts mean that it is necessary to explicitly model the gas dynamics (‘rules of
thumb’ do not always work, and are sometimes misleading as Blair has pointed out
[24]). A 1D gas dynamics model such as the one developed here is an indispensable
tool for exploring design options for engine charging.
185
Chapter 9 Summary and conclusion
This thesis has covered a lot of territory, which can be conveniently grouped into
two main topics – namely free-piston engine technology developments and general
engine modelling. This wide subject range reflects the particular needs of the
Pempek project for a wide range of both results analysis and predictive modelling.
All free-piston engineer researchers develop some kind of model or models to test
their designs – indeed it is fairly easy to construct a basic free-piston engine cylinder
model to determine operating frequency and the like. Some of the models used by
others are very advanced, such as detailed in-cylinder CFD, detailed chemical
kinetics, and in some cases 1D gas dynamics programs.
The modelling tool that this author developed for the Pempek engine is
comparatively simple, as it does not attempt to model detailed in-cylinder processes,
however it lays great emphasis on accurately determining the inlet and exhaust flows
by modelling the gas dynamics of the entire inlet and exhaust duct. The model is
simple enough to be easy to set up, and fast to execute on a computer, which allows
effective design to be carried out on a large range of cylinder charging options. It is
also generally applicable to a wide range of other gas dynamic processes.
This author believes that too few free-piston researchers understand the potential of
gas dynamics to influence engine charging. (See section 1.5.4) Free-piston engines
(especially 2-stroke ones) are ideally suited to take advantage of gas dynamics
because they are essentially constant speed machines.
The following sections summarise the specific findings for the Pempek project, the
scope and usefulness of the model, the unique contributions made in the field of
modelling, and finally a few areas of suggested further research.
Chapter 9 Summary and conclusion 186
9.1 Findings for the Pempek project
This author’s association with Pempek Systems has lasted for about four years, and
during that time a range of engine performance issues were examined. Early work
included a simple free-piston engine cylinder model from which estimates of exhaust
blowdown pressure were used in the design of an updated exhaust valve actuator.
Other early work involved a detailed moving mesh CFD cylinder model for testing
the influence of inlet valve placement on cylinder scavenging. (see the movie
‘points’ in Appendix XV for an example). This work suggested that about 8%
residual exhaust would be retained, and that valve placement here did not greatly
affect scavenging efficiency, however exhaust pressure and inlet valve lift had to be
estimated. Following this work, the integral compressor was analysed and
recommended to be re-designed for lower pressure in order to reduce power
consumption and inlet air heating.
Findings that relate directly to the engine model described in this thesis are as
follows:
Passive inlet valve trajectory – the simulated valve trajectory showed that re
seating velocities were as high as 3.5m/s and that the valve typically opened at least
twice each cycle. Simulations showed that a weaker return spring could be used.
Lower compressor pressure possible – simulations showed that enough air
delivery could be made with a lower compressor pressure, especially if a tuned
exhaust pipe assisted in evacuating the cylinder after blowdown.
Exhaust tuning beneficial – simulations of a port admission version of the engine
showed that no compressor was necessary for cylinder charging if a carefully tuned
exhaust pipe was used. That is, according to modelling results, 2-stroke engines
operating in a narrow speed range can be charged by using the gas dynamic energy
of exhaust flows.
Chapter 9 Summary and conclusion 187
9.2 The model, its scope and usefulness
The engine model described in this thesis was developed to model the Pempek
engine. It is actually a collection of independent models for
thermodynamic control volumes
gas properties and chemical equilibrium
1D unsteady gas dynamic flow
1D multi body dynamics
various miscellaneous models such as heat transfer, combustion rate and
blowby
These models can be applied to a wide range of processes, not just internal
combustion engines. A graphical user interface was created to allow the user to
easily create, assemble and edit engine models. The multi body dynamics model is a
special requirement of free-piston engine modelling, and the correct handling of
collisions is particularly important for the Pempek engine with its passive inlet
valve; however other models for motion can be prescribed by the user.
Since it is important to easily visualise simulation data, two special graphing utilities
were created. One facilitates the plotting of time series data, while the other allows
gas dynamic animations, which give the engine designer clear insight into the gas
dynamic processes within the engine. Several of these animations are included with
this thesis and are listed in Appendix XV.
The 1D gas dynamics model allows variations in gas properties, is energy and mass
conservative, and includes an extensive treatment of boundary flows. It is ideally
suited to high performance two stroke engines with strong wave action, thermal and
gas property discontinuities and diverging and converging ducts. It provides for an
engine model which is comprehensive enough to the capture the significant physics
of the engine cycle (especially engine breathing), while being simple and fast enough
to allow rapid evaluation of design options such as
optimal valve timing
optimal valve sizing
optimal inlet and exhaust duct shape
necessary inlet charge pressurisation etc.
delivery ratio
Chapter 9 Summary and conclusion 188
9.3 Unique contributions to modelling art
During the course of developing the engine model, the author devised and
implemented several new methods.
Chemical equilibrium code – This code was based on the popular method of
equilibrium constants, however equation solution and the detailed code was written
entirely by the author. The solution is efficient and reliable, taking an average of 5
iterations of a 4x4 matrix division to reach sufficient accuracy. It can handle gas
mixtures without atoms of carbon, hydrogen or nitrogen - only oxygen is
compulsory. (See section 3.3)
Second order pressure wave interpolation method – Wave action gas dynamics
models are typically first order accurate since they use linear interpolation of
pressure wave values between mesh nodes. A simple but effective second order
interpolation method was devised and implemented. (See section 4.3.1) Smearing is
significantly reduced compared to the first order method, and Courant numbers
slightly above unity can be safely used.
Heat transfer and mass conservation in 1D model– The heat transfer
implementation is different to the original method from Queens University, Belfast.
It performs well in the Rayleigh flow test. Furthermore, the method includes a
provision for full mass conservation, which is notable, since wave action methods
usually are not fully mass conservative. (See section 4.3.2)
Duct re-meshing – This procedure allows the mesh spacing of individual ducts to be
changed mid simulation to better suit the prevailing flow conditions, and maintain a
Courant number closer to unity. (See section 4.3.3)
Moving mesh – this capability is useful for some situations where ducts deform,
such as a moving piston problem. This capability was added to the original method
by a simple change of reference frame for the travelling pressure waves (See section
4.4.8)
Fluid property interpolation method (with variable mixing) – The original
method used the upstream cell values for gas properties at nodal flow. This results
in unavoidable diffusion of thermal and mixture discontinuities. A reasonably
simple interpolation method was devised and implemented which allows the user to
set various rates of mixing and can maintain sharp gas property discontinuities. (See
section 4.5.2)
Chapter 9 Summary and conclusion 189
Duct friction and heat transfer rate model – This author’s analysis of
Kirkpatrick’s single shot data [68], as well as consideration of the process of
boundary layer development in impulsive type flows, found that a constant
coefficient of friction was more suitable than the more commonly used Blasius
formula.. (See section 5.1.1) Following similar reasoning, a heat transfer model was
devised which depended on turbulence kinetic energy in the bulk flow to account for
highly elevated heat transfer in instances of strong turbulence. (See section 5.1.4)
Diffusing flow model – The original gas dynamics method used the momentum
equation approximation to model diffusing flow pressure recovery. Initially, a
relationship which included flow unsteadiness [56] was developed, but in the
absence of showing a clear advantage, has been abandoned in favour of a simpler
steady relationship. (See section 5.3).
Chapter 9 Summary and conclusion 190
9.4 Contribution to free-piston engine research
This body of work, taken as a whole, offers two main contributions specific to the
field of free-piston research. The first is that it demonstration of the importance of
gas dynamics effects in two-stroke free-piston engines. Section 1.5.4 reviews the
existing references to gas dynamics modelling in recent free piston literature.
Further, details of recent free piston projects are given in Appendix I. These sections
show that not many free piston engine researchers have attempted to model or
exploit gas dynamic effects. Section 2.3, especially Figure 2-10 shows that the
Pempek engine with ordinary exhaust tube is clearly affected by gas dynamic effects
given the pressure fluctuations in the cylinder during charging of up to 0.5 bar. Even
in the un-optimised state this represents a fluctuation in cylinder trapped mass of +-
25%. The introduction to Chapter 4 demonstrates that from a purely theoretical
point of view, neglecting gas dynamics can lead to large errors in calculated mass
flows.
Conventional engine modellers have long been convinced of the power of gas
dynamics to help or harm engine performance. The modelling in this thesis of
possible design options for the Pempek engine (Chapter 8) shows that gas dynamics
effects have the power to dramatically improve the charging efficiency, and if
necessary, to charge the cylinder without assistance of intake blowers or
compressors. The wide range of possible design options that gas dynamic effects
introduces are discussed in the conclusion to Chapter 8 (section 8.4)
The second contribution to the field of free piston engine research is a thorough
literature survey of recent free piston projects (section 1.2) This survey is followed
by a discussion of some central free piston engine concerns such as piston motion
control (section 1.3). Finally, a more in depth report of the Pempek engine design
and experimental results (Chapter 2) allows those interested in the field to learn from
Pempek’s experience.
Chapter 9 Summary and conclusion 191
9.5 Further Work
During the course of the work, several avenues for further research were identified,
but because of time constraints were not pursued. With regard to unsteady 1D
modelling:
Multi-pipe junction model - A good multi-pipe junction model is an important part
of a fully capable engine model – though it was unnecessary for the existing Pempek
engine since the inlet and exhaust systems were independent for each cylinder.
There is quite a bit of literature on this topic, and perhaps understandably (given the
simplifications inherent in a 1D model), none of the models reported so far have
been able to reproduce experimental results for a full range of cases.
Diffusing flow model – This author believes there is more scope for investigating
this very typical flow mode. Evidence seems to point to reduced flow separation for
the case of impulsive flow compared to steady flows, but some reports are
conflicting. This flow mode is also important in multi-pipe junction flows, where
gas flows around quite sharp bends and experiences a loss of total pressure.
Unsteady duct heat transfer – A relatively simple model for unsteady heat transfer
based on modelled turbulence energy was quickly developed for this work. More
work needs to be done in examining the physical basis for such a model, and
validating and refining it against a range of experimental cases.
***
With regard to free-piston engines, it would be useful to do a study of the feasibility
of a contactless piston. (See section 8.2.1) Port inlet and/or exhaust has great
advantage over valve systems in terms of simplicity, zero power consumption and in
some configurations, uni-flow scavenging. Normally however, cylinder ports can
cause accelerated piston ring wear and increased oil consumption. An engine with a
ringless piston, no-contact ‘seal’ could be almost maintenance free. The free-piston
engine with its linear motion and low side forces is an ideal candidate for such a
technology. Probably the main difficulty with the concept is potential un-balanced
side pressure forces on the piston due to piston-cylinder eccentricity during the high
pressure part of the cycle. Thus, a careful piston dynamics study is required which
models the three dimensional piston-cylinder crevice flow and the resulting pressure
forces on the piston.
192
. . . .
193
PUBLICATIONS
Greg Gibbes and Guang Hong, Improving the accuracy of a 1D Gas Dynamics
Model, Proceedings of the 17th Australasia Fluid Mechanics Conference, Auckland,
New Zealand, 5-9 December 2010.
Greg Gibbes and Guang Hong, A General Boundary Solution Method for 1D Gas
Dynamic Models, Proceedings of the 17th Australasia Fluid Mechanics Conference,
Auckland, New Zealand, 5-9 December 2010.
Greg Gibbes and Guang Hong, A model for pressure recovery of flows in expanding
ducts for 1D gas dynamics simulations, Proceedings of the 15th Asia Pacific
Automotive Engineering Conference (APAC15), Hanoi, Vietnam, 26-28 October,
2009.
Greg P. Gibbes and Guang Hong, Numerical modelling the dynamics of a piston-
mounted passive inlet poppet valve, SAE 2007-32-0099, JSAE 20076599,
Proceedings of the 13th Small Engine Technology Conference (SETC), Niigata
Japan, 30 October-1 November 2007.
194
. . . .
195
REFERENCES
[1] AVL BOOST, 2012, AVL, https://www.avl.com/boost1 Acessed 8/7/2012
[2] GT-Power, 2012, Gamma Technologies, http://www.gtisoft.com/ Acessed
30/6/2012
[3] Lotus Engine Simulation, 2012, Lotus,
http://www.lotusfiles.com/engineering/lesoft/LES_insert.pdf Acessed 8/7/2012
[4] Virtual 2-stroke, 2012, Optimum Power Technology, http://www.optimum-
power.com/Virtual%202-Stroke.htm Acessed 8/7/2012
[5] WAVE, 2012, Ricardo, http://www.ricardo.com/en-GB/What-we-
do/Software/Products/WAVE/ Acessed 8/7/2012
[6] Achten, P. A. J., 1994, A Review of Free Piston Engine Concepts, SAE
Paper 941776
[7] Achten, P. A. J., Oever, J. P. J., Potma, J., and Vael, G. E. M., 2000,
Horesepower with Brains: The Design of the CHIRION Free Piston Engine, SAE,
Paper 2000-01-2545
[8] Aichlmayr, H. T., 2002, Design Considerations, Modeling, and Analysis of
Micro-Homogeneous Charge Compression Ignition Combustion Free-Piston
Engines, PhD, THE UNIVERSITY OF MINNESOTA.
[9] Aichlmayr, H. T., 2002, Micro-Engine Work at the University of Minnesota,
http://www.menet.umn.edu/~haich/engine/index.html.
[10] Aichlmayr, H. T., Kittelson, D. B., and Zachariah, M. R., 2002, Miniature
free-piston homogeneous charge compression ignition engine-compressor concept--
Part I: performance estimation and design considerations unique to small
dimensions, Chemical Engineering Science, 57(19)4161-4171.
[11] Aichlmayr, H. T., Kittelson, D. B., and Zachariah, M. R., 2002, Miniature
free-piston homogeneous charge compression ignition engine-compressor concept--
Part II: modeling HCCI combustion in small scales with detailed homogeneous gas
phase chemical kinetics, Chemical Engineering Science, 57(19)4173-4186.
[12] Aichlmayr, H. T., Kittelson, D. B., and Zachariah, M. R., 2003, Micro-
HCCI combustion: experimental characterization and development of a detailed
chemical kinetic model with coupled piston motion, Combustion and Flame,
135(3)227-248.
[13] Anderson, J. D., 1990, Modern Compressible Flow, McGraw-Hill
Publishing Company.
References 196
[14] Annand, W. J. D., 1963, Heat Transfer in the Cylinders of a Reciprocating
Internal Combustion Engine, Proc. Instn. Mech. Engrs.(177)973.
[15] Ariffin, A. K., Mohamed, N. A. N., and Fonna, S., 2006, Simulation of Free
Piston Linear Engine Motion with Different Intake and Exhaust Port Positions,
Jurnal Kejuruteraan(18)97-106.
[16] Arshad, W. M., Thelin, P., Backstrom, T., and Sadarangani, C., 2004, Use
of transverse-flux machines in a free-piston generator, Industry Applications, IEEE
Transactions on, 40(4)1092-1100.
[17] Aziz, A.-R. A., Rangkuti, C., and Baharom, M., 2004, Dynamics Of CNG-
Fuelled Free-Piston Linear Generator Engine, NGV 2004, Buenos Aires, Argentina,
International Association for Natural Gas Vehicles,
[18] Bannister, F. K., and Mucklow, G. F., 1948, Wave action following sudden
release of compressed gas from a cylinder, Proceedings of the Institution of
Mechanical Engineers, vol. 159269-300.
[19] Bassett, M. D., Pearson, R. J., and Fleming, N. P., 2003, A Multi-pipe
Junction Model for One-dimensional Gas-dynamic Simulations, SAE, Paper 2003-
01-0370
[20] Bauer, W.-D., Wenisch, J., and Heywood, J. B., 1998, Averaged and time-
resolved heat transfer of steady and pulsating entry flow in intake manifold of a
spark-ignition engine, International Journal of Heat and Fluid Flow, 191-9.
[21] Benson, R. S., 1982, The Thermodynamics and Gas Dynamics of Internal
Combustion Engines Volume 1, Oxford University Press, Walton Street, Oxford.
[22] Benson, R. S., Garg, R. D., and Woollatt, D., 1964, A Numerical Solution of
Unsteady Flow Problems, Int. J. Mech. Sci., vol. 6117.
[23] Blair, G. P., 1991, An Alternative Method for the Prediction of Unsteady
Gas Flow Through the Internal Combustion Engine, SAE, Paper 911850
[24] Blair, G. P., 1999, Design and simulation of Four-Stroke Engines, Society of
Automotive Engineers, Warrendale PA U.S.A.
[25] Blair, G. P., Callender, E., and Mackey, D. O., 2001, Maps of Discharge
Coefficients for Valves, Ports and Throttles, Small Engine Technology Conference
and Exhibition, Pisa, Italy, SAE International, Paper 2001-01-1798
[26] Blair, G. P., Kirkpatrick, S. J., and Fleck, R., 1995, Experimental Validation
of a 1d Modelling Code for a Pipe Containing Gas of Varying Properties, SAE,
Paper 950275
References 197
[27] Blair, G. P., Kirkpatrick, S. J., Mackey, D. O., and Fleck, R., 1995,
Experimental Validation of 1-D Modelling Codes for a Pipe System Containing Area
Discontinuities, SAE, Paper 950276
[28] Blair, G. P., Lau, H. B., Cartwright, A., Raghunathan, B. D., and Mackey, D.
O., 1995, Coefficients of Discharge At the Apertures of Engines, SAE, Paper
952138
[29] Blair, G. P., Mcburney, D., Mcdonald, P., Mckernan, P., and Fleck, R.,
1998, Some Fundamental Aspects of the Discharge Coefficients of Cylinder Porting
and Ducting Restrictions, SAE, Paper 980764
[30] Borman, G. L., and Ragland, K. W., 1998, Combustion Engineering,
McGraw-Hill.
[31] Bosman, C. B., and Goldsborough, S. S., 2008, Developing a 10cc Single-
Valve, Reverse Uniflow, 2S Engine, SAE, ed., SAE, Paper 2008-01-0953
[32] Brusstar, M., Charles Gray, J., Jaffri, K., McCarthy, P., and Pomerleau, M.,
2005, Design, Development and Testing of Multi-Cylinder Hydraulic Free-Piston
Engines, SAE 2005 World Congress & Exhibition, Detroit, MI, USA, Paper 2005-
01-1167
[33] Carter, D., and Wechner, E., 2003, The Free Piston Power Pack:
Sustainable Power for Hybrid Electric Vehicles, SAE, Paper 2003-01-3277
[34] Cawthorne, W. R., Famouri, P., Jingdong, C., Clark, N. N., McDaniel, T. I.,
Atkinson, R. J., Nandkumar, S., Atkinson, C. M., and Petreanu, S., 1999,
Development of a linear alternator-engine for hybrid electric vehicle applications,
Vehicular Technology, IEEE Transactions on, 48(6)1797-1802.
[35] Chalet, D., Chesse, P., Hetet, J.-F., and Tauzia, X., 2009, Inflow boundary
condition for one-dimensional gas dynamics simulation code of internal combustion
engine manifolds, Proc. IMechE Part D: J. Automobile Engineering, Vol. 223.
[36] Chalet, D., Chesse, P., and Hetet, J. F., 2008, Boundary conditions
modelling of one-dimensional gas flows in an internal combustion engine,
International Journal of Engine Research, 9.
[37] Chalet, D., Chesse, P., Tauzia, X., and Hetet, J.-F., 2006, Comparison of
Different Methods for the Determination of Pressure Wave in the Inlet and Exhaust
Systems of Internal Combustion Engine, SAE, Paper 2006-01-1542
[38] Chase, M. W., Davies, C. A., J.R. Downey, J., Frurip, D. J., McDonald, R.
A., and Syverud, A. N., 1985, JANAF Thermochemical Tables: Third Edition,
American Chemical Society and the American Institute of Physics for the National
Bureau of Standards.
References 198
[39] Cosic, A., Sadarangani, C., and Carlson, F., 2005, A novel Concept of a
Transverse Flux Linear Free-Piston Generator, Linear Drives for Industry
Applicatons (LDIA 2005), Hyogo, Japan,
[40] Cosic, A., Sadarangani, C., and Timmerman, J., 2008, Design and
Manufacturing of a Linear Transverse Flux Permanent Magnet Machines, Proc.
Industry Applications Society Annual Meeting, 2008. IAS '08. IEEE
[41] Depcik, C., and Assanis, D., 2002, A Universal Heat Transfer Correlation
for Intake and Exhaust Flows in an Spark-Ignition Internal Combustion Engine,
SAE, Paper 2002-01-0372
[42] Depcik, C., van Leer, B., and Assanis, D., 2004, The Numerical Simulation
of Variable-Property Reacting-Gas Dynamics: New Insights and Validation,
Numerical Heat Transfer, Part A: Applications, 47(1)27 - 56.
[43] Dittus, P. W., and Boelter, L. M. K., 1930, Heat Transfer in Automobile
Radiators of the Tubular Type, University of California Berkeley Publications on
Engineering, 2(13)443-461.
[44] Earnshaw, S., 1860, On the Mathematical Theory of Sound, Philosophical
Transactions of the Royal Society of London, 150133-148.
[45] EPSRC, 2005, Zero constraint free piston energy converter grant details,
http://gow.epsrc.ac.uk/ViewGrant.aspx?GrantRef=GR/S97514/01
[46] Erickson, W. D., and Prabhu, R. K., 1986, Rapid computation of chemical
equilibrium composition - an application to hydrocarbon combustion, Journal Name:
A.I.Ch.E. J.; (United States); Journal Volume: 32,OSTI ID: 7137953 p.^pp.
Medium: X; Size: Pages: 1079-1087.
[47] Faiz, J., Rezaeealam, B., and Yamada, S., 2006, Reciprocating flux-
concentrated induction generator for free-piston generator, Magnetics, IEEE
Transactions on, 42(9)2172-2178.
[48] Famouri, P., Cawthorne, W. R., Clark, N., Nandkumar, S., Atkinson, C.,
Atkinson, R., McDaniel, T., and Petreanu, S., 1999, Design and testing of a novel
linear alternator and engine system for remote electrical power generation, Proc.
Power Engineering Society 1999 Winter Meeting, IEEE
[49] Ferguson, C. R., and Kirkpatrick, A. T., 2001, Internal Combustion Engines
- Applied Thermosciences, John Wiley & Sons.
[50] Foley, L., Douglas, R., McCullough, G., and Fleck, R., 1998, One-
Dimensional Mass and Energy Transport Using a Modified Mesh Method, SAE,
Paper 982049
References 199
[51] Fredriksson, J., Bergman, M., Golovitchev, V. I., and Denbratt, I., 2006,
Modeling the Effect of Injection Schedule Change on Free Piston Engine
Operation,SAE Paper 2006-01-0449 p.^pp.
[52] Fredriksson, J., and Denbratt, I., 2004, Simulation of a Two-Stroke Free
Piston Engine, SAE 2004 Spring Fuels & Lubricants Conference, Toulouse, SAE,
Paper 2004-01-1871
[53] Galindo, J., Lujan, J. M., Serrano, R. R., Dolz, V., and Guilain, S., 2006,
Description of a heat transfer model suitable to calculate transient processes of
turbocharged diesel engines with one-dimensional gas-dynamic codes, Applied
Thermal Engineering, 2666-76.
[54] Gault, R. I., Thornhill, D. J., and Fleck, R., 2008, Alternative method to
evaluate discharge coefficients. Part 2: case study, Proc. IMechE Vol. 223 Part C: J.
Mechanical Engineering Science.
[55] Gibbes, G., and Hong, G., 2007, Numerically Modeling the Dynamics of a
Piston-Mounted Passive Inlet Poppet Valve, Small Engine Technology Conference,
Niigata, Japan, SAE, Paper 2007-32-0099
[56] Gibbes, G., and Hong, G., 2009, A Model for Pressure Recovery of Flows in
Expanding Ducts for 1D Gas Dynamic Simulations, Asia Pacific Automotive
Conference, Hanoi, Vietnam, SAE,
[57] Goertz, M., and Peng, L., 2000, Free Piston Engine Its Application and
Optimistaion,SAE Paper 2000-01-0996 p.^pp.
[58] Goldsborough, S., and Van Blarigan, P., 2003, Optimizing the Scavenging
System for a Two-Stroke Cycle, Free Piston Engine for High Efficiency and Low
Emissions: a Computational Approach, 2003 SAE World Congress, Detroit,
Michigan, SAE, SAE Paper 2003-01-0001
[59] Goldsborough, S. S., and Van Blarigan, P., 1999, A Numerical Study of a
Free Piston IC Engine Operating on Homogeneous Charge Compression Ignition
Combustion, SAE Paper 1999-01-0619
[60] Graef, M., Treffinger, P., Pohl, S.-E., and Rinderknecht, F., 2007,
Investigation of a High efficient Free Piston Linear Generator with variable Stroke
and variable Compression Ratio, The World Electric Vehicle Association Journal,
1116-120.
[61] Hew Wooi, P., Arof, H., and Wijono, 2006, Design of a Permanent Magnet
Linear Generator, Proc. Strategic Technology, The 1st International Forum on
[62] Heywood, J. B., 1988, Internal Combustion Engine Fundamentals,
McGraw-Hill.
References 200
[63] Hibi, A., and Ito, T., 2004, Fundamental test results of a hydraulic free
piston internal combustion engine, Proceedings of the Institution of Mechanical
Engineers, 218(10).
[64] Houdyschell, D., 2000, A Diesel Two-Stroke Linear Engine, Master of
Science in Mechanical Engineering, West Virginia University, Morgantown, West
Virginia.
[65] INNAS, 2010, INNAS free piston engine site,
http://www.innas.com/CFPE.html.
[66] Johansen, T. A., Egeland, O., Johannessen, E. A., and Kvamsdal, R., 2002,
Free-Piston Diesel Engine Timing and Control - Towards Electronic Cam- and
Crankshaft, IEEE Transactions on Control Systems Technology, 10(2)177-190.
[67] Johansen, T. A., Egeland, O., Johannessen, E. A., and Kvamsdal, R., 2003,
Dynamics and Control of a Free-Piston Diesel Engine, Journal of Dynamic Systems,
Measurement, and Control, 125(3)468-474.
[68] Kirkpatrick, S. J., 1994, An Evaluation of One-Dimensional Simulation
Techniques for Predicting Unsteady Gas Flow in Engine Ducting, Doctor of
Philosophy, The Queens University of Belfast.
[69] Kirkpatrick, S. J., Blair, G. P., Fleck, R., and McMullan, R. K., 1994,
Experimental Evaluation of 1-D Computer Codes for the Simulation of Unsteady
Gas Flow Through Engines - a First Phase, SAE, Paper 941685
[70] Kleemann, A. P., Dabadie, J.-C., and Henriot, S., 2004, Computational
Design Studies for a High-Efficiency and Low-Emissions Free Piston Engine
Prototype, SAE, Paper 2004-01-2928
[71] Klein, M., 2004, A Specific Heat Ratio Model and Compression Ratio
Estimation.pdf, Linköping University, Sweden, Linköping.
[72] Krieger, R. B., and Borman, G. L., 1967, The computation of apparent heat
release for internal combustion engines, ASME,
[73] Laboratory for the Design and Control of Energetic Systems, V. U., 2011,
Free-Piston Engine Compressor Homepage,
http://www.vanderbilt.edu/dces/fpc.html.
[74] Larmi, M., Isaksson, S., Tikkanen, S., and Lammila, M., 2001, Performance
Simulation of a Compression Ignition Free Piston Engine, SAE 2001 World
Congress, Deltroit, Michigan, SAE, Paper 2001-01-0280
[75] Li, Q., Xiao, J., and Huang, Z., 2008, Simulation of a Two-Stroke Free-
Piston Engine for Electrical Power Generation, Energy & Fuels, 22(5)3443-3449.
References 201
[76] Mao, J., Zuo, Z., Li, W., and Feng, H., 2010, Multi-dimensional scavenging
analysis of a free-piston linear alternator based on numerical simulation.pdf,
Applied Energy.
[77] Matlab, 2010, Acessed
[78] Max, E., 2005, FPEC, Free Piston Energy Converter, 21st Int. Electric
Vehicle Symposium, Monte Carlo, Monaco,
[79] Cemical Equilibrium with Applications (CEAgui), Glenn Research Center,
http://www.grc.nasa.gov/WWW/CEAWeb/ceaHome.htm Acessed
[80] Mikalsen, R., and Roskilly, A. P., 2007, The design and simulation of a
two-stroke free-piston compression ignition engine for electrical power generation,
Applied Thermal Engineering.
[81] Mikalsen, R., and Roskilly, A. P., 2007, Performance simulation of a spark
ignited free-piston engine generator, Applied Thermal Engineering.
[82] Mikalsen, R., and Roskilly, A. P., 2007, A review of free-piston engine
history and applications, Applied Thermal Engineering(27 )2339–2352.
[83] Mikalsen, R., and Roskilly, A. P., 2008, A computational study of free-
piston diesel engine combustion, Applied Energy.
[84] Mikalsen, R., and Roskilly, A. P., 2008, Coupled dynamic-multidimensional
modelling of free-piston engine combustion, Applied Energy.
[85] Mikalsen, R., and Roskilly, A. P., 2009, The control of a free-piston engine
generator. Part 1: fundamental analyses, Applied Energy.
[86] Mikalsen, R., and Roskilly, A. P., 2009, The control of a free-piston engine
generator. Part 2: Engine dynamics and piston motion control, Applied Energy.
[87] Mikalsen, R., and Roskilly, A. P., 2009, The fuel efficiency and exhaust gas
emissions of a low heat rejection free-piston diesel engine, Proc. IMechE Vol. 223
Part A: J. Power and Energy, 223.
[88] Mikalsen, R., and Roskilly, A. P., 2009, Predictive piston motion control in
a free-piston internal combustion engine, Applied Energy.
[89] Mohammed, N. A. N., Ariffin, A. K., and Fonna, S., 2006, Simulation of a
Two-Stroke Spark Ignition Free Piston Linear Engine Motion, Jurnal Teknologi,
44(A)27-40.
[90] Munson, B. R., Young, D. F., and Okiishi, T. H., 2002, Fundamentals of
Fluid Mechanics, John Wiley & Sons.
[91] Nandkumar, S., 1998, Two-Stroke Linear Engine, Master of Science In
Mechanical Engineering, West Virginia University, Morgantown, West Virginia.
References 202
[92] Nanni, D., Romoli, M., and Rossi, R., 2007, 1D/3D Comparison of Flow
Field Simulations Inside an Exhaust-Type Duct, SAE, Paper 2007-01-1398
[93] Newcastle_University, 2010, Free Piston Engine Project Page, Newcastle
University, Newcastle upon Tyne, , http://www.free-piston.eu/.
[94] Olikara, C., and Borman, G., 1975, A Computer Progrram for Calculating
Properties of Equilibrium Combustion Products sith Some Applications to I.C.
Engines, SAE, Paper 750468
[95] Pearson, R., Bassett, M., Virr, P., Lever, S., and Early, A., 2006, Exhaust
system gas dynamics in internal Combustion Engines, ASME Internal Combustion
Engine Division 2006 Spring Technical Conference, Aachen, Germany, ICES2006-
1444
[96] Pempek-Systems, 2009, Free Piston Power Project main page,
www.freepistonpower.com.
[97] Petreanu, S., 2001, Conceptual Analysis of a Four-Strok Linear Engine,
Doctor of Philosophy in Mechanical Engineering, West Virginia University,
Morgantown, West Virginia.
[98] Rao, S. S., 2002, Applied Numerical Methods for Engineers and Scientists,
Prentice Mall.
[99] Rashidi, M., 1998, Calculation of equilibrium composition in combustion
products, Applied thermal Engineering, 18(3-4)103-109.
[100] Rerkpreedapong, D., 1999, Field Analysis and Design of a Moving Iron
Linear Alternator for Use with Linear Engine, Master of Science, West Virginia
University, Morgantown.
[101] Riemann, B., 1860, Ueber die Fortpflanzung ebener Luftwellen von
endlicher Schwingungsweite., Abhandlungen der Königlichen Gesellschaft der
Wissenschaften zu Göttingen, vol. 8.
[102] Riofrio, J. A., and Barth, E. J., 2008, Experimental Assessment of a Free
Elastic-Piston Engine Compressor with Separated Combustion Chamber, ASME
Symposium on Fluid Power and Motion Control (FPMC 2008), Bath, U K, ASME,
[103] Roth, K., Sobiesiak, A., Robertson, L. A., and Yates, S., 2002, In-Cylinder
Pressure Measurements With Optical Fiber and Piezoelectric Pressure Transducers,
SAE, Paper 2002-01-0745
[104] Selamet, E. E., Selamet, A., and Novak, J. M., 2004, Predicting the
chemical species in spark-igintion engiens, Energy, 29449-465.
References 203
[105] Shoukry, E. F., 2003, Numerical Simulation for Parametric Study of a Two-
Stroke Compression Ignition Direct Injection Linear Engine, Doctor of Philosophy
in Mechanical Engineering, West Virginia University, Morgantown, West Virginia.
[106] Smith, G. L., D.Erickson, W., and Eastwood, M. R., 1967, Equations for the
rapid machine computation of equilibrium composition of air and derivatives for
flow field calculations, Langley Research Center, NASA.
[107] Sonntag, R. E., Borgnakke, C., and Van Wylen, G. J., 2003, Fundamentals
of Termodynamics, John Wiley & Sons, New York.
[108] Teh, K.-Y., Miller, S. L., and Edwards, C. F., 2008, Thermodynamic
requirements for maximum internal combustion engine cycle efficiency. Part 1:
optimal combustion strategy, International Journal of Engine Research, 9(6).
[109] Teh, K.-Y., Miller, S. L., and Edwards, C. F., 2008, Thermodynamic
requirements for maximum internal combustion engine cycle efficiency. Part 2: work
extraction and reactant preparation strategies, International Journal of Engine
Research, 9(6).
[110] Thornhill, D., Currie, T., Fleck, R., and Chatfield, G., 2006, Particle image
velocimetry investigation of a finite amplitude pressure wave, Experiments in Fluids,
40400-404.
[111] Thornhill, D., Li, H., Fleck, R., and Cuningham, G., 2006, Modeling of
unseady gas-dynamic flow in a pipe and at its exit using CFD, Proc. IMechE Part D:
J. Automobile Engineering, 220.
[112] Tian, L., Kittelson, D. B., and Durfee, W. K., 2009, Miniature HCCI Free-
Piston Engine Compressor, SAE, Paper 2009-32-0176
[113] Tikkanen, S., Lammila, M., Herranen, M., and Vilenius, M., 2000, First
Cycles of the Dual Hydraulic Free Piston Engine, International Off-Highway &
Powerplant Congress & Exposition, Milwaukee, Wisconsin, SAE, Paper 2000-01-
2546
[114] Tikkanen, S., and Vilenius, M., 2006, Control of dual hydraulic free piston
engine, International Journal of Vehicle Autonomous Systems, 4-120.
[115] Tóth-Nagy, C., 2004, Linear Engine Development for Series Hybrid Electric
Vehicles, Doctor of Philosophy in Mechanical Engineering, West Virginia
University, Morgantown, West Virginia.
[116] Van Blarigan, P., 1999, Homogenous Charge Compression Ignition With a
Free Piston: A New Approach to Ideal Otto Cycle Performance, Proc. U.S DOE
Hydrogen Program Review, Department of Energy,NREL/CP-570-26938
References 204
[117] Van Blarigan, P., 2001, Advanced Internal Combustion Engine Research,
Proc. 2000 DOE Hydrogen Program Review,NREL/CP-570-28890
[118] Van Blarigan, P., 2002, Advanced Internal Combustion Electrical
Generator, Proceedings of the 2002 U.S. DOE Hydrogen Program
Review,NREL/CP-610-32405 p.^pp.
[119] Van Blarigan, P., 2009, Free Piston Engine, Sandia National Laboratories.
[120] Van Blarigan, P., 2010, Free Piston Engine, Sandia National Laboratories.
[121] Van Blarigan, P., Paradiso, N., and Goldsborough, S., 1998, Homogeneous
Charge Compression Ignition with a Free Piston: A New Approach to Ideal Otto
Cycle Performance, SAE, Paper 982484
[122] Wang, J., West, M., Howe, D., Parra, H. Z.-D. L., and Arshad, W. M., 2007,
Design and Experimental Verification of a Linear Permanent Magnet Generator for
a Free-Piston Energy Converter, IEEE TRANSACTIONS ON ENERGY
CONVERSION, 22(2).
[123] Willhite, J. A., and Barth, E. J., 2009, Reducing Piston Mass in a Free
Piston Engine Compressor by Exploiting the Inertance of a Liquid Piston, ASME
Dynamic Systems and Control Conference and Bath/ASME Symposium on Fluid
Power & Motion Control (FPMC), ASME,
[124] Winterbone, D. E., and Pearson, R. J., 1999, Theory of Engine Manifold
Design: Wave Action Methods for IC Engineers., Professional Engineering
Publishing Ltd, Suffolk, UK.
[125] Woo, Y., Lee, Y., and Lee, Y., 2009, The performance characteristics of a
hydrogen-fuelled free piston internal combustion engine and linear generator
system, International Journal of Low-Carbon Technologies, 4(1)11.
[126] Wu, W., Yuan, S., Hu, J., and Jing, C., 2009, Design approach for single
piston hydraulic free piston diesel engines, Frontiers of Mechanical Engineering in
China, 4(4).
[127] Xiao, Y., 1989, Thermodynamic properties for engine combustion
simulations, College of Engineering, University of Michigan,
http://hdl.handle.net/2027.42/8365.
[128] Xu, Z., and Chang, S., 2009, Modelling and control of an internal
combustion linear generator integrated power system, Int. J. Modelling,
Identification and Control, 7(4)398-404.
[129] Yong, C., Riofrio, J. A., and Barth, E. J., 2008, Modeling and Control of a
Free-Liquid-Piston Engine Compressor, ASME Symposium on Fluid Power and
Motion Control (FPMC 2008), Bath, U K, ASME,
References 205
[130] Zeng, P., and Assanis, D., 2004, Unsteady Convective Heat Transfer
Modeling and Application to Engine Intake Manifolds, Proc. ASME International
Mechanical Engineering Congress and RD&D Expo, SAE,
[131] Zheng, P., Chen, A., Thelin, P., Arshad, W. M., and Sadarangani, C., 2007,
Research on a tubular Longitudinal Flux PM Linear Generator Used for Free-
Piston Energy Converter, IEEE Transactions on Magnetics, 43(1).
[132] Zhu, Y., and Reitz, R. D., 1999, A 1-D gas dynamics code for subsonic and
supersonic flows applied to predict EGR levels in a heavy-duty diesel engine, Int. J.
of Vehicle Design, Vol. 22(3/4)227-252.
[133] Zulkifli, S. A., Karsiti, M. N., and Aziz, A., 2008, Starting of a free-piston
linear engine-generator by mechanical resonance and rectangular current
commutation, Proc. Vehicle Power and Propulsion Conference, 2008. VPPC '08.
IEEE
206
. . . .
207
APPENDICES
Appendix I Review of recent free-piston engine projects 209
Appendix II Specific heats of a reacting mixture 235
Appendix III Further applications of the energy equation 237
Appendix IV Tables of Thermodynamic Properties 241
Appendix V Method for Calculating Chemical Equilibrium of Combustion
Products 247
Appendix VI Derivation of fundamental one dimensional unsteady gas equation
255
Appendix VII Derivation of Boundary Flow Equations 259
Appendix VIII Model Data Structures 263
Appendix IX 2nd Order Interpolation – further details 269
Appendix X Re-meshing Criteria and Method 271
Appendix XI Single Shot Experiments Cross Reference 273
Appendix XII Derivation of normal shock equations 275
Appendix XIII Rayleigh and Fanno Flow 281
Appendix XIV Graphical user interface screen shots 285
Appendix XV Table of contents of data CD 291
208
. . . .
209
Appendix I
Review of recent free-piston engine projects
This review will focus on contemporary projects published from about the year 2000
onward. These recent projects are briefly summarised according to their supporting
institutions. Design details and results of particular relevance to this thesis are
emphasised. Thus more attention is paid to electric free-piston engine projects (as
opposed to hydraulic), projects with experimental results, and with similar
mechanical design issues to the engine studied in this project. A cursory summary of
linear electric machine projects is given, though mainly to provide context for the
current state of the field.
I-1 Overview of early Free-piston engines
The first commercially successful free-piston engines were designed in the 1920’s.
The early history of free-piston engines is well explored by Aichlmayr [8], and the
following details of early engines are summarised from his thesis. Other worthwhile
reviews can be found in [6] and [82].
Prolific inventor Raúl Pateras Pescara developed a number of designs for free-piston
air compressors beginning in 1922. At around the same time the German company
Junkers began developing similar air compressors, which were used to provide
compressed air for torpedo launch tubes on German submarines. Both the Pescara
and Junkers style machines were opposed piston designs utilising a synchronising
mechanism to ensure the pistons maintained perfect opposed motion. This design
was convenient for the purpose of gas exchange, since one piston uncovered exhaust
ports, while the other uncovered inlet ports. Air compression was generally multi
stage and was achieved with various integrated pistons. The Junkers design was
highly successful and following the Second World War captured Junkers free-piston
compressors were distributed to various U. S. firms, Universities, and laboratories
for evaluation and testing. Figure I-1 shows a cutaway view of a Junkers air
compressor.
Appendix I Review of recent free-piston engine projects 210
Figure I-1 Junkers four-stage free-piston air compressor [8]
The next major phase of free-piston engine technology was free-piston gas
generators (gasifiers). These engines produced high pressure, high temperature gas
which was used to supply a gas turbine for power extraction. At a time when gas
turbine technology was in its infancy, a free-piston gasifier gave higher combustion
temperatures and pressures, while allowing the gas turbine to run on relatively low
temperature partially expanded gas. French company SIGMA produced the GS-34
in 1944 based on Pescara’s designs. This engine had a final power output of around
600kW and was produced for large scale applications such as electrical power
generation and marine power. About 600 of these engines were eventually produced
and saw successful service for more than two decades. Figure I-2 shows the SIGMA
GS-34. General Motors developed a similar machine - the GM-14, but with limited
success, and tested a much smaller machine – the GMR 4-4 ’Hyprex – for
automotive application, but this too was unsuccessful. Ford also experimented with
free-piston gasifiers without commercial success.
Appendix I Review of recent free-piston engine projects 211
Figure I-2 Partial cut-away diagram of the SIGMA Type GS-34 Free-Piston
Gasifier [8]
Appendix I Review of recent free-piston engine projects 212
I-2 Hydraulic free-piston engines
I-2-1 INNAS
In the early 1990’s, Dutch company Innas began developing a single ended direct
injected compression ignition hydraulic free-piston engine. The technical features
and measured performance of the INNAS free-piston engine (named Chiron) is
described by Achten in [7]. It has a nominal maximum power output of 17kW to a
hydraulic rail pressure of 260-320 bar. Output power control is achieved by pulse
pause modulation. Indicated efficiency is around 50% and net effective efficiency is
above 32% for the 25-100% load range. The Chiron engine as shown in Figure I-3 is
a mature prototype with demonstrably good piston control and performance.
However there has been little further published information on this engine since the
year 2000. NOAX continues to market and develop hydraulic hybrid vehicle
systems for INNAS. The INNAS website [65] provides some basic information.
INNAS and NOAX have been involved in the EU supported Free-piston Energy
Converter (FEPC) project since late 2002.
Figure I-3 INNAS “Chiron” hydraulic free-piston engine [7]
I-2-2 Tampere and Helsinki Universities of Technology
In 1999 Tampere University of Technology designed and built a prototype opposed
cylinder, hydraulic free-piston engine based on patents held by Sampower Ltd Oy.
Appendix I Review of recent free-piston engine projects 213
This engine is shown in Figure I-4. Experimental results are reported by Tikkanen et
al. in [113]. Similar to the INNAS engine (above), this is a loop scavenged two-
stroke (using cylinder wall ports) with direct injection and compression ignition
however, it is an opposed cylinder machine. Measured output power was about
11kW with a net efficiency of about 20%.
Figure I-4 Tampere University of Technology hydraulic free-piston engine
prototype “Emma2” [113]
Combustion was found to be fast enough (about 2.5ms) and reliable. Some
asymmetry in the combustion pressures was evident, and this was attributed to un-
equal fuel injection, asymmetric pump design and manufacturing tolerances in the
cylinder assemblies.
Helsinki University of Technology undertook a simulation effort to analyse and
improve the engine [74].
A revised prototype (Figure I-5) was constructed with largely similar dimensions to
the initial prototype. The power was doubled by lightening the piston assembly
(increasing the operating frequency), slightly increasing the effective stroke, and
improvements to the hydraulic pump, scavenging and control system. Tikkanen and
Vilenius [114] report a control system designed for operating this machine. The
controller was tested on a Matlab/Simulink model of the engine and showed good
ability to control compression ratio, provided that the rate of change in load level
was kept within certain limits. No experimental results were reported.
Appendix I Review of recent free-piston engine projects 214
Figure I-5 Third Generation prototype engine [114]
I-2-3 Toyohashi University of Technology
Hibi and Ito [63] report the latest experimental results from a long running hydraulic
free-piston engine project. The engine described is an opposed piston, direct
injected diesel compression ignition machine with a newly designed hydraulic circuit
and a peak power of almost 5kW. The combustion piston diameters are each 100mm
they each have an effective compression stroke of a little over 100mm. The piston
assemblies weigh 4.16kg each. The machine is intended to implement an engine-off
system utilising a hydraulic accumulator. The machine operates in single cycles,
allowing power output to be modulated, while maintaining optimum combustion
conditions. The maximum frequency reported was 7.25Hz (though judging by the
piston motion this could probably be increased significantly). The paper describes
the hydraulic circuit and control of the new engine in some detail, and gives
experimental hydraulic and gas pressures throughout an engine cycle. Triggering of
each engine cycle (from the piston’s holding positions at the outward position) is
activated when hydraulic pressure in the main accumulator falls below a certain
threshold.
Results showed that the net efficiency of the machine was substantially constant over
the full range of load level, staying at 31 1 percent, which though unexceptional,
is significant since the part load efficiency does not fall. A cross section of the
machine is shown in Figure I-6.
Appendix I Review of recent free-piston engine projects 215
Figure I-6 Cutaway view of the Toyohashi university of Technology hydraulic
free-piston engine [63]
I-2-4 U.S. EPA, National Vehicle and Fuel Emissions Laboratory
The U.S. EPA engaged in an intensive free-piston engine development program as
part of their hydraulic hybrid vehicle program. Engine and powertrain development
company FEV were also contracted to this project and brought existing expertise
[57]. Brusstar et al. [32] describe the design and operating performance of two
prototype hydraulic free-piston engines,. The engines were both based on an
opposed cylinder module. One engine was a two cylinder two stroke version with a
rated output of 22kW, while the other was a six cylinder four stroke with a rated
output of 54kW. Both engines were compression ignition, utilising high pressure
direct injection or optional port fuel injection for HCCI operation. Scavenging in the
two-stroke was uni-flow via inlet cylinder ports and cam or hydraulic driven exhaust
valves. Valving in the four stroke version was hydraulics driven. Significantly, both
engines were operated for extended periods and the four stroke version demonstrated
an estimated net efficiency of 34-39% over much of the load range. Figure I-7
shows a sketch of the six cylinder four stroke, demonstrating reasonably flat
packaging.
Appendix I Review of recent free-piston engine projects 216
Figure I-7 EPA six-cylinder four-stroke FPE [32]
I-2-5 Beijing Institute of Technology
Wu et al. describe the design and operation of a single ended diesel hydraulic free-
piston engine [126]. The hydraulic circuit is described in some detail and
preliminary experimental results are given for motored operation. Various design
issues are discussed such as the need for fast acting, high flow hydraulic valves, and
a control system capable of dealing with misfires. Contrary to the control method of
the EPA machine (above), which used controllable check valves to vary hydraulic
load at arbitrary operating pressures, this engine will vary the fuel quantity to
balance against the hydraulic load at different operating pressures. A numerical
model of the machine including piston dynamics, combustion cylinder
thermodynamics and hydraulic system dynamics is also described and simulation
results are presented.
Appendix I Review of recent free-piston engine projects 217
I-3 Electric free-piston engines
I-3-1 University of West Virginia
West Virginia University have had a long running free-piston engine project that
attracted funding from the US army and has been the subject of several post-graduate
theses. Nandkumar [91] describes the early work on a spark ignition prototype build
around chainsaw engine pistons which had a bore of 36.5mm and stroke of about
40mm. Gasoline mixed with lubricating oil was injected near the cylinder ports.
The engine was initially tested with a rudimentary friction brake on the translator
assembly to extract mechanical work. It seems that fuelling and throttling levels
were set prior to individual test runs, and were not dynamically controlled during a
run. Due to the shape of the combustion chamber, the maximum achievable
compression ratio was about 8:1 since any further travel of the piston would result in
it striking the head. To prevent over-stroke ignition timing was radically advanced.
Somewhat surprisingly, the engine could run stably from zero load right up to the
stalling load, without any piston control, due to the self-regulating nature of the early
ignition, which automatically reduced thermodynamic efficiency when there was
little load on the translator. Presumably the problem of limited compression ratio
was only exacerbated under throttled operation. A permanent magnet alternator was
later used in place of the friction brake as a passive (uncontrolled) load.
Based on the experimental results and numerical modelling, a second generation
prototype was planned and built. Houdyschell [64] describes the initial design of the
engine which used significantly larger pistons (75mm diameter) and a total stroke of
about 70mm. Fuel was direct injection diesel or kerosene, though only kerosene was
used since the engine required a spark ignition stroke to start. The new engine was
designed to operate at higher frequency and compression ratio. It is shown in Figure
I-8. Due to regular catastrophic failure of various parts of the engine, Tóth-Nagy
[115] writes “The engine development was a continuous improving and redesigning
process”. Nevertheless, sufficient test results were obtained to do a fairly in-depth
analysis of the characteristics of the engine.
Appendix I Review of recent free-piston engine projects 218
Figure I-8 West Virginia University, second generation linear engine prototype
(2-stroke) [115]
Again, no active load control was used on this engine, and it was initially run with
the same friction brake concept as the first prototype. It thus suffered like its
predecessor from over-stroke at low load, and peak cylinder pressures as high as
30Mpa were recorded. A closed loop fuel control was proposed to allow fuelling to
better match the applied load.
Petreanu [97] simulated a concept opposed four cylinder, four-stroke design, in an
attempt to circumvent the problems he considers inherent in the two stroke concept.
He states “The main problem of two-stroke cylinder processes is that they offer low
fuel efficiency due to the imperfect scavenging processes.” The proposed layout of
the four stroke engine is shown in Figure I-9.
Appendix I Review of recent free-piston engine projects 219
Figure I-9 West Virginia University, four stroke concept [97]
Shoukry [105] reports more numerical modelling work on the two stroke concept,
finding among other things that shorter strokes lead to higher power operation but
reduced efficiency.
I-3-2 Sandia National Laboratories
The Sandia free-piston project grew out of research into hydrogen and high
compression ratio piston engines. In [121] van Blarigan reports the results of
combustion experiments using a rapid compression expansion machine (RCEM).
Fuels tested were propane, natural gas, hydrogen, methanol, npentane, hexane, n-
heptane, and isooctane. Ideal case indicated thermal efficiency of natural gas and
propane were the best at about 56% and did not increase significantly at higher
compression ratios. The good performance of these fuels is attributed to the very
fast (single step) burn characteristic giving effectively constant volume combustion.
The equivalence ratio used in the tests varied slightly depending on the fuel but all
were around 0.35 (lean). The main conclusions of the study are as follows:
Compression ratios as high as 30:1 are possible before the onset of auto-
ignition – though fuel, initial charge temperature and piston speed influence
the timing of combustion.
A number of fuels were capable of very rapid combustion – approximating
the constant volume “ideal”.
Appendix I Review of recent free-piston engine projects 220
Over-compression due to premature combustion did not appear to adversely
affect cycle efficiency, but did tend to produce higher NOx emissions.
NOx could be controlled by lean operation or dilution with exhaust gas.
Different fuels had very different combustion rates. In order to ensure
complete combustion of some fuels, increased CR, increased initial charge
temp or slower piston motion could be employed.
Van Blarigan suggests that the main reason for the lack of expected efficiency
improvement at higher CR may be time loss (ie non constant volume combustion).
He dismisses increased heat loss as a primary cause since several tests with an
uncoated piston crown and cylinder head showed decreased efficiency of only 5%.
He did not however seem to consider the potential detrimental influence of increased
crevice volume flows at higher compression ratios.
Scott Goldsborough [58] reports an extensive CFD modelling project which he
conducted for his PhD thesis at the Colorado State University on the planned free-
piston engine shown in Figure I-10, a HCCI, port fuelled engine.
Figure I-10 First Sandia Free-piston Linear Alternator concept [116]
Goldsbrough’s 3D simulation gave best net cycle efficiency at 52%. While the
efficiency figures are useful for comparison between cases, Goldsborough
apparently did not account for blowby and crevice volume flow.
Soon after this during modelling work with the NASA [118] it was found that a
turbocharger would provide sufficient energy for pressurising the inlet air to drive
scavenging.
Soon after the completion of Goldsbrough’s modelling efforts on an opposed
cylinder, uni-flow layout (using inlet ports and exhaust valves) an opposed piston
layout was proposed having a single central combustion chamber as shown in Figure
I-11.
Appendix I Review of recent free-piston engine projects 221
Figure I-11 Sandia opposed piston layout [119]
The chief benefits of the opposed piston layout are inherent mechanical balance
when the pistons are synchronised, and the opportunity to utilise a central
combustion chamber with inlet and exhaust ports at opposite ends. This gives uni-
flow scavenging without the need for any poppet valves. On the other hand,
accurate piston synchronisation is essential, the overall length of the engine is
increased over a single piston concept, two extra bounce cylinders are required, as
are two generators. It is not clear how much of the design of this prototype is
intended to embody a final commercial engine layout. Features of this engine design
are:
twin linear alternators connected electrically in parallel to effect the piston
synchronisation. The load on the alternator coils is resistive at present (not
actively switched), and is constant at about 15kW in total. This is
considered necessary to maintain the synchronisation.
High pressure air injection into bounce chambers to achieve piston motion
during un-powered and low powered cycles.
Port fuel injection (low delivery ratio essential to avoid fuel short circuiting)
The compression ratio control is intriguing. Under “motoring” conditions, over
15kW of compressed air supply is required to drive the piston oscillation against the
constant generator load (the linear generator is passive). If fuel injection is
increased, increasing combustion energy allows the vent pressure in the bounce
chamber to be increased, presumably to the point where no extra air injection is
required. One potential criticism of the concept is losses incurred in the bounce
chambers form heat transfer, gas leakage, valve flow. Careful design should be able
to minimise these.
Appendix I Review of recent free-piston engine projects 222
Figure I-12 Sandia bounce chamber detail showing compressed air injection
valves on the left and vent ports on the right [119]
The aims of the project are [119]
To study the effects of continuous operation (i.e. gas exchange) on indicated
thermal efficiency and emissions at high compression ratios (~20-40:1)
Concept validation of passively coupling the opposed free-pistons via the
linear alternators connected to a common load to maintain piston
synchronization
Proof of principle of electronic variable compression ratio control
Figure I-13 shows the complete opposed piston prototype, awaiting final installation
of the bounce chamber injectors and latch mechanisms. It will provide a platform
for testing the free-piston operating envelope over varying levels of boost,
equivalence ratio, and alternative fuels.
Appendix I Review of recent free-piston engine projects 223
Figure I-13 Sandia Opposed Piston Free-piston Engine [120]
Current project partners are Los Alamos National Laboratory (engine and generator
modelling), General Motors/University of Michigan Collaborative Research
Laboratory (engine modelling) and Stanford University.
Figure I-14 gives a timeline of the project. The project has also included a
substantial effort in linear generator design.
Figure I-14 Timeline of Sandia free-piston engine project [119]
I-3-3 FPEC
A European project consortium with funding from the Swedish Energy Agency and
the European Commission has been developing a free-piston engine since 2002
called the “Free-piston Energy Converter” (FPEC). A completed prototype is shown
in Figure I-15. According to [70] the engine has a bore of 102mm and a nominal
Appendix I Review of recent free-piston engine projects 224
stroke of about 127mm, and will operate at a high compression ratio of 26:1 with
50% EGR. [78] provides a good summary of the results of the project as of 2005,
just prior to the commencement of experimental testing. The engine is opposed
cylinder, diesel direct injected (to operate with early injection for HCCI-like
combustion). Chalmers University of Technology, Sweden have conducted
extensive modelling of the combustion and combustion related aspects – see for
example [51]. A tubular permanent magnet machine was designed by Sheffield
University and constructed by ABB Corporate Research, Sweden. [122] The electric
machine was predicted to be capable of operating at about 30kW continuously with
an efficiency of 93%. In parallel, the Royal Institute of Technology Stockholm,
Sweden (KTH) have designed and built an alternative electric machine [40] using an
unusual transverse flux design. French research organisation IFP conducted initial
modelling work to verify and the expected performance of the scavenging and
determine optimum valve timing and port layout [70]. Dutch companies INNAS and
NOAX with hydraulic free-piston engine experience are project members. Volvo
Technology Corporation, Sweden is also a project partner.
Figure I-15 FPEC prototype electric generator – opposed cylinder, two-stroke
diesel with pnumatic exhaust valves [78]
I-3-4 University of Newcastle upon the Tyne
The University of Newcastle upon Tyne began a free-piston engine project in 1999
[93]. In the last couple of years, Mikalsen and Roskilly have published numerous
papers detailing the modelling of a proposed two stroke, single ended electric free-
piston engine [80].
Appendix I Review of recent free-piston engine projects 225
I-3-5 Pempek
Australian industrial electronics company Pempek Systems Pty. Ltd. developed and
tested a compact two stroke, opposed cylinder engine[96]. The details of this project
are described in Chapter 2.
I-3-6 Korea Institute of Energy
An opposed cylinder spark ignition engine prototype built by the Korea Institute of
Energy is described in [125]. The engine was tested with both CNG and Hydrogen
fuel. Variability in the stroke caused some instability in operation. The control
method is not disclosed, though it seems that the linear generator is a passive
machine, and fuel mass is the control parameter. A revised design is proposed which
replaces the original ported loop scavenging with a uni-flow system in which poppet
valves in the cylinder head evacuate the exhaust gases and the generator is modified
to provide integral supercharging of the inlet air.
Figure I-16 Korea Institute of Energy opposed cylinder engine [125]
I-3-7 Malaysian Ministry of Science, Technology and Environment
A free-piston engine project funded by the Ministry of Science, Technology and
Environment, Malaysia is reported in [17]. This engine is a single ended spark
ignition machine with air cushion bounce chamber and was built using the piston and
cylinder head of a two stroke motorcycle engine in similar fashion to the University
of West Virginia machines (described above). The engine was operated with no load
and the piston motion analysed using a high speed camera. A related project [89]
describes a numerical model of a similar engine utilising a linear electric machine for
power extraction. Work on an appropriate linear electric machine for the free-piston
engine is reported from the University of Malaya [61] and the Universiti Teknologi
PETRONAS, Malaysia [133]
Appendix I Review of recent free-piston engine projects 226
I-3-8 Loughborough/Sheffield/Lotus
Loughborough University partnered with Sheffield University and Lotus
Engineering Ltd. in 2005 under a UK government EPSRC grant to develop a four
stroke free-piston engine [45].
I-3-9 Shanghai Jiao Tong University
Li et al. [75] report a numerical investigation of an opposed cylinder machine with
linear alternator with special emphasis on HCCI operation. A chemical kinetics
model (Chemkin/Senkin) was used for the in-cylinder combustion process which
was assumed homogenous, adiabatic and perfectly sealed. The force characteristic
of the linear alternator was analysed using the Finite Element Method, and this
compared favourably with simple analytical expressions which were incorporated in
the engine model. The alternator was modelled with an uncontrolled resistive load.
The piston dynamics model was implemented with Matlab/Simulink. The study
traced the relationships between generator load, compression ratio and equivalence
ratio. The shorter residence time at TDC compared to a cranked engine was noted,
with possible benefits being reduced NOx formation and reduced heat transfer. It is
unclear if any closed loop control is implemented in the model (such as equivalence
ratio control), though given the inherent damping characteristic of a resistive load,
closed loop control may not be necessary if the combustion strength is fairly
constant.
I-3-10 Nanjing University of Science and Technology
The Nanjing University of Science and Technology [128] report the design and
modelling of a novel four-stroke, single ended, spark ignition free-piston engine as
sketched in Figure I-17. Springs attached to the mover provide the energy storage
for exhaust the intake strokes, while the large energy requirement of the compression
stroke is provided by the linear electric machine during this stroke. The control
system uses electromagnetic force as the main control variable, and treats fuel
delivery as an essentially fixed input. Needless to say, the electric machine must
allow bidirectional energy transfer. Simulation results of piston dynamics show the
viability of this approach.
Appendix I Review of recent free-piston engine projects 227
Figure I-17 Four-stroke free-piston engine concept by Nanjing University of
Science and Technology [128]
I-3-11 German Aerospace Center
A project at the German Aerospace Center report preliminary modelling and testing
of various components of a free-piston linear generator [60]. The proposed engine is
a single ended machine with electromagnetic inlet and exhaust valves. Notably, a
permanent magnet linear generator has been constructed and tested, showing good
performance due to effective cooling. Efficiency of the engine’s gas spring has been
measured at 90-95% depending on engine speed. Heat loss to walls was found to be
the primary loss mode with blowby accounting for only 10% of loss. Piston motion
control is based on generator load control. The prototype engine is shown in Figure
I-18.
Figure I-18 German Aerospace Center free-piston prototype on test bench [60]
Appendix I Review of recent free-piston engine projects 228
I-4 Summary of linear generator projects
A suitable linear generator is one of the most challenging design problems facing the
development of electric free-piston engines. High efficiency and high power-to-
weight ratio must be combined in the one machine, which must also be able to
withstand the mechanical shock of repeated high mover accelerations. Many of the
electric free-piston engine projects cited above are devoting significant effort in this
area. While a detailed survey of this field is beyond the scope of this thesis, a brief
(and no doubt, incomplete) summary of some linear generator projects is given
below:
1999 As part of the West Virginia University (WVU) Project, [34, 48] describe the
design and testing of a permanent magnet linear alternator. Maximum power output
at 23Hz was 316W. The alternator was tested with the opposed cylinder engine and
was connected to an adjustable, purely resistive load. A cross section of the
alternator is shown in Figure I-19. Rerkpreedapong [100] describes the design and
modelling of an alternative machine using a moving iron topology. This has the
advantage of placing the permanent magnets in the stator, thus allowing improved
cooling and protection from the high acceleration loads of the mover. However the
mass of the mover is increased compared to moving magnet designs, so power
density is lower due to reduced mover frequency. Predicted efficiency was 79%.
Figure I-19 Linear alternator built at WVU [34]
2006 Faiz et al. [47] from the University of Tehran describe the modelling and
testing of a self-excited reciprocating generator with flux concentrators. The merits
of the design are a lightweight mover, low maintenance, and robust structure.
Conducting plates are inserted into the stator slots to achieve flux concentration by
eddy currents. The presence of the plates increases the performance of the engine
Appendix I Review of recent free-piston engine projects 229
significantly; however the efficiency is not as good as permanent magnet type
generators.
The Sandia free-piston engine project designed and built two similar linear
alternators. One was developed in-house (Figure I-20). The stator was made from
1600 ground tapered laminations, and each of the 25 coils contained 78 turns of
square copper wire. Nominal power output is 40kW at 94% efficiency. A second
alternator was made by Magnaquench. The stator of this machine was made from
powdered iron in an adhesive matrix (Figure I-21). The actual performance of these
machines is unknown, however it appears that the magnaquench machine is being
used in the latest dual alternator machine described in [119].
Figure I-20 Sandia linear alternator design [117]
Figure I-21 Magnaquench linear alternator stator [117]
Work at the University of Sheffield on a machine for the Free-piston Energy
Converter (FPEC) has culminated in the machine decribed in [122]. The machine
has three phases and employs a quasi-Halbach permanent magnet arrangement on
the mover and since no back iron is required, the permanent magnets are mounted
Appendix I Review of recent free-piston engine projects 230
directly on a non-magnetic support tube. The machine has a high power density, low
cogging torque and has a simplified manufacturing process. The stator is made in
modules with flat laminations.
Figure I-22 University of Sheffield FPEC linear permanent magnet generator
[122]
An alternative transverse flux machine has been developed at the Royal Institute of
Technology, Sweden [16, 39, 40] as shown in Figure I-23. Simulation work on a
more conventional longitudinal flux machine is also reported [131].
Figure I-23 Royal Institute of Technology Stockholm, Sweden transverse flux
permanent magnet generator [40]
Appendix I Review of recent free-piston engine projects 231
[133] mentions a conventional surface magnet generator developed by Universiti
Teknologi PETRONAS (UTP) in collaboration with University Malaya (UM) and
University Kebangsaan Malaysia (UKM). [61] describes an alternative machine
from the UM using a quasi Halback permanent magnet arrangement to eliminate the
need for back iron.
The German Aerospace Center have constructed a prototype linear generator which
is mentioned briefly in [60].
Figure I-24 German Aerospace Center linear generator on the test stand [60]
I-5 Other free-piston engines
I-5-1 University of Minnesota
A numerical and experimental program at the University of Minnesota investigated
the viability of miniature free-piston engine with a power output of 10W. The
rationale behind the investigation was the superior energy density of liquid
hydrocarbon fuels compared with battery technology. Among the technical issues
present with miniature combustion engines, the work here focused on micro-
combustion [10], in particular HCCI combustion [11]. The characteristics of HCCI
combustion make it particularly advantageous for miniature engines. Since the
charge is ignited by compression heating, no ignition system is required. The
presence of multiple ignition points and well mixed charge ensure rapid, even
combustion, minimising combustion chamber wall quenching effects and allowing
high engine speeds. The use of the free-piston concept allows variable compression
ratio to be used as a combustion phasing control. Single shot experiments [12]
confirmed that a large alkane (heptane) at a lean equivalence ratio (0.25) could be
reliably combusted in a space 3mm in diameter and 0.3mm in length, with sufficient
Appendix I Review of recent free-piston engine projects 232
speed to produce virtually constant volume combustion. Subsequent analysis of the
experimental data using a numerical model allowed piston blowby to be quantified.
Further details of the project (which ended in 2002) can be found at the project
website [9].
More recently, the centre for Compact and Efficient Fluid Power (CCEFP) have
reported design, modelling and preliminary experimental work on a miniature HCCI
air compressor free-piston engine [112]. The concept design is shown in Figure I-
25.
Figure I-25 Design concept for a miniature HCCI free-piston engine [112]
I-5-2 Kvaerner ASA and Norwegian University of Science and
Technology
Kvaerner ASA built a free-piston gasifier engine with exhaust gas turbine for power
extraction in a development effort aimed at marine applications as an alternative to
both gas turbines and traditional diesel engines. It consists of multiple single piston
units connected to common intake and exhaust manifolds. The engine takes some
design cues from the commercially successful Pescara/SIGMA free-piston gasifiers.
The Kvaerner prototype is a single piston unit with a nominal power output of 1MW.
A photograph of the engine rig is shown in Figure I-26. Control focuses on BDC
and TDC position control using fuel mass and bounce chamber air mass respectively
as control parameters. Frequency can be varied slightly during operation by
changing the stroke length, providing a mechanism of keeping multiple cylinders
properly synchronised. Johansen et al. [66] describe the salient details of the engine,
Appendix I Review of recent free-piston engine projects 233
test rig and control system, and give some experimental results which demonstrate
that the control system performs well in maintaining proper piston motion control
over the load range. In a companion paper, Johansen et al. [67] describe a
mathematical model of the engine dynamics which is used to derive an engine
control system structure.
Figure I-26 Photo of Kvaerner test cylinder unit
I-5-3 Marquette University
Bosman and Goldsbourough build on Goldsbourough’s previous work with Sandia
to do preliminary scavenging and combustion modelling of a simple two-stroke free-
piston engine with opposed cylinders designed for driving hydraulic or pneumatic
loads [31]. The engine uses a passive inlet valve in the cylinder head and exhaust
ports at the bottom section of the cylinder. Small engines are envisaged; however
the design could be scaled up for automotive applications. Initial scavenging
modelling reveals high residual exhaust fraction due to entrainment to the lee of the
inlet valve, and non-optimal valve and port timing.
Appendix I Review of recent free-piston engine projects 234
I-5-4 Vanderbilt University
A free-piston air compressor for applications of human scale robotics has been
reported from the Laboratory for the Design and Control of Energetic Systems at
Vanderbilt University. A slug of fluid contained in a cylinder and on each end by a
flexible membrane forms a “liquid piston” with perfect sealing and negligible
friction. A combustion chamber on one end drives the liquid piston to pump air into
a pressurised reservoir on the other end [102, 129]. Subsequent experimental and
modelling work revealed the need to slow down the dynamics of the system, so
rather than increase the piston mass, a high inertance piston was devised [123].
Figure I-27 Liquid piston air compressor prototype (high inertance model)
Vanderbilt University [73]
235
Appendix II
Specific heats of a reacting mixture
Specific heat is the amount of heat per unit mass required to raise the temperature of
a substance by one degree. Two distinct specific heats are defined, one at constant
pressure and the other at constant volume.
However, for reacting mixtures (in chemical equilibrium), there is another important
intensive property to consider, and this is the relative species fractions.
Traditionally, specific heat for a reacting mixture is defined for the case of shifting
(chemical) equilibrium – that is, the specific heat that would be measured if an
experiment was conducted on the mixture in question. The derivation is as follows.
The internal energy of a mixture of species is given by
Differentiating WRT to temperature gives
Likewise, for a mixture of ideal gases, the enthalpy of a mixture is given by
Differentiating WRT to temperature gives
Thus for a reacting mixture the true specific heats are defined as
Appendix II Specific heats of a reacting mixture 236
Then the ratio of specific heats is
and
It is useful however, to make an alternative definition for specific heat, where the
chemical composition of the fluid is held constant (frozen). In this case, the
chemical reaction term is zero, and the frozen specific heats become.
The ratio of frozen specific heats is
Then
Frozen specific heats are used throughout this thesis for reacting mixtures. This is
because of the choice to model reacting mixtures explicitly, with each species
contributing to the energy balance of a mixture. All chemical reaction effects (be
they dissociation in shifting equilibrium or fuel oxidisation etc.) are written into the
energy equation explicitly, rather than being subsumed into the specific heat.
237
Appendix III
Further applications of the energy equation
The form of the energy equation (3-2) used in this engine model was chosen for its
ability to conveniently and accurately model the thermodynamics of arbitrary
reacting mixtures. The following section offers alternative forms of the equation
which may be useful for certain applications. Most significantly, it is shown how the
energy equation can be used “in reverse” to analyse cylinder pressure data for
combustion, heat transfer and blowby characteristics.
The energy equation for an open control volume containing several chemical species
is given by equation (3-2).
Expanding the flow enthalpy term the energy equation becomes.
Using and separating the species mass flow out of the term
and relocating it to the flow enthalpy term yields
(III-1)
where the now represents the change in energy due to chemical reactions
only. If the specific heat of each species is reasonably constant over the
temperature range between the bulk volume temperature and the flow temperatures
then internal energy terms of the mass flows can be substituted away.
Appendix III Further applications of the energy equation 238
(III-2)
Evaluating heat release
Equation (III-2) can be written in terms of pressure so as to estimate the chemical
reaction heat release from experimental engine data.
The ideal gas equation is:
Differentiating with respect to time gives
Given that
and
Then
And differentiating WRT time
So the ideal gas equation in rate form becomes
(III-3)
Appendix III Further applications of the energy equation 239
So assuming the cylinder can be modelled as an ideal gas, equation (III-3) is used to
substitute for in Equation (III-2) and bringing the chemical energy change to
the LHS gives
Writing in terms of the frozen specific heat ratio and neglecting the second flow
enthalpy term gives.
Note that for an exothermic reaction, the chemical energy change is negative. Note
also mass flow and heat transfer is positive into the cylinder. P and can be
obtained from pressure transducer data. Likewise the piston motion must be known
to find V and . The chemical heat release rate can be numerically integrated
to determine the cumulative total heat release. If mass flows (eg fuel injection,
crevice flows, blowby, valve leakage) and heat transfer are neglected then the
resulting heat release is termed apparent.
In heat release calculations the term is often neglected. Thus the third term
on the RHS of equation (III-4) is not usually considered. Figure III- shows that for a
mixture in shifting chemical equilibrium significant variation in mixture R does not
happen until the mixture is heated above about 2500K. Also, the gas constant of the
un-reacted mixture (279 J/kg/K in this case) or pure air (287 J/kg/K) is not too
dissimilar to that of the burned state.
(III-4)
Appendix III Further applications of the energy equation 240
Figure III-1 Variation of R with temperature for products of combustion in
chemical equilibrium
280
300
320
340
360
380
400
0 1000 2000 3000 4000 5000
1 bar 10 bar
100 bar
0T (K)
R (J
/kg/
K)
Mixture made from 75 parts N2, 15 parts O2 and 5 parts n-octane (C8H18) by mass
241
Appendix IV
Tables of Thermodynamic Properties
Table IV-1 Enthalpy of some combustion products Enthalpy Reference Temp=298.15 Absolute enthalpy (J/mol) Standard state Pressure = 0.1Mpa
Source JANAF Thermochemical Tables (3rd Edition) T (K) CO CO2 H H2 H2O N N2 NO O O2 OH M 28.0104 44.0098 1.00794 2.01588 18.01528 14.0067 28.0134 30.0061 15.9994 31.9988 17.00734
100 -116296 -399978 213880 -5468 -248441 468564 -5768 84218 244655 -5779 32848 200 -113385 -396936 215959 -2774 -245108 470643 -2857 87340 246987 -2868 36011 300 -110473 -393453 218037 53 -241764 472721 54 90346 249214 54 39042 400 -107551 -389519 220116 2959 -238374 474800 2971 93331 251380 3025 42022 500 -104596 -385217 222195 5882 -234901 476879 5911 96350 253516 6084 44979 600 -101585 -380615 224273 8811 -231325 478957 8894 99435 255635 9244 47930 700 -98504 -375768 226352 11749 -227634 481036 11937 102598 257743 12499 50889 800 -95350 -370716 228430 14702 -223824 483114 15046 105839 259844 15835 53867 900 -92126 -365492 230509 17676 -219888 485193 18223 109149 261940 19241 56875
1000 -88837 -360125 232588 20680 -215826 487272 21463 112520 264033 22703 59922 1100 -85492 -354638 234666 23719 -211635 489350 24760 115944 266123 26212 63011 1200 -82097 -349049 236745 26797 -207320 491429 28109 119411 268212 29761 66147 1300 -78659 -343374 238823 29918 -202884 493507 31503 122917 270299 33344 69329 1400 -75184 -337626 240902 33082 -198333 495586 34936 126455 272385 36957 72556 1500 -71677 -331817 242981 36290 -193675 497665 38405 130020 274469 40599 75826 1600 -68142 -325953 245059 39541 -188918 499743 41904 133610 276554 44266 79138 1700 -64582 -320042 247138 42835 -184068 501822 45429 137220 278637 47958 82489 1800 -61001 -314091 249216 46169 -179133 503901 48978 140848 280720 51673 85876 1900 -57401 -308103 251295 49541 -174120 505979 52548 144492 282803 55413 89297 2000 -53783 -302083 253374 52951 -169036 508058 56137 148150 284886 59175 92749 2100 -50151 -296034 255452 56397 -163885 510137 59742 151821 286969 62961 96230 2200 -46506 -289960 257531 59876 -158673 512217 63361 155503 289051 66769 99739 2300 -42848 -283862 259609 63387 -153405 514297 66995 159195 291135 70600 103272 2400 -39179 -277743 261688 66928 -148085 516378 70640 162897 293218 74453 106828 2500 -35500 -271605 263767 70498 -142718 518460 74296 166607 295303 78328 110406 2600 -31812 -265449 265845 74096 -137306 520543 77963 170325 297389 82224 114004 2700 -28116 -259276 267924 77720 -131853 522628 81639 174050 299476 86141 117620 2800 -24411 -253089 270003 81369 -126362 524716 85323 177782 301564 90079 121254 2900 -20700 -246886 272081 85043 -120836 526807 89015 181520 303654 94036 124905 3000 -16981 -240670 274160 88740 -115277 528901 92715 185264 305747 98013 128571 3100 -13256 -234441 276238 92460 -109687 531000 96421 189013 307842 102009 132252 3200 -9526 -228201 278317 96202 -104069 533103 100134 192768 309940 106023 135947 3300 -5789 -221949 280396 99966 -98423 535213 103852 196527 312040 110054 139654 3400 -2047 -215686 282474 103750 -92753 537329 107577 200291 314144 114102 143374 3500 1700 -209413 284553 107555 -87058 539452 111306 204059 316252 118165 147106 3600 5451 -203129 286631 111380 -81341 541585 115041 207832 318363 122245 150850 3700 9208 -196836 288710 115224 -75604 543726 118781 211609 320478 126339 154604 3800 12969 -190533 290789 119089 -69846 545877 122525 215389 322597 130447 158368 3900 16734 -184221 292867 122972 -64069 548040 126274 219174 324720 134569 162142 4000 20504 -177900 294946 126874 -58274 550215 130027 222962 326848 138705 165926 4100 24277 -171571 297024 130795 -52463 552402 133784 226753 328981 142854 169719 4200 28055 -165232 299103 134734 -46635 554603 137545 230548 331118 147015 173521 4300 31836 -158885 301182 138692 -40792 556819 141309 234347 333260 151188 177332 4400 35621 -152531 303260 142667 -34934 559050 145078 238148 335407 155374 181151 4500 39410 -146168 305339 146660 -29062 561297 148850 241953 337559 159572 184978 4600 43202 -139797 307417 150670 -23176 563560 152625 245760 339716 163783 188814 4700 46998 -133419 309496 154698 -17278 565841 156405 249571 341878 168005 192657 4800 50797 -127033 311575 158741 -11368 568140 160187 253385 344045 172240 196508 4900 54599 -120640 313653 162801 -5446 570458 163973 257201 346218 176488 200367 5000 58405 -114239 315732 166876 487 572794 167763 261021 348395 180749 204233 5100 62214 -107831 317810 170967 6432 575150 171556 264843 350578 185023 208107 5200 66026 -101413 319889 175071 12389 577526 175352 268668 352765 189311 211988 5300 69841 -94987 321968 179190 18358 579921 179152 272495 354957 193614 215876 5400 73660 -88551 324046 183322 24338 582338 182955 276325 357155 197933 219771 5500 77481 -82106 326125 187465 30331 584775 186761 280158 359357 202267 223674 5600 81306 -75652 328203 191621 36335 587233 190571 283994 361564 206618 227584 5700 85134 -69188 330282 195787 42351 589711 194384 287832 363775 210987 231501 5800 88965 -62716 332361 199963 48378 592211 198201 291672 365991 215375 235425 5900 92799 -56234 334439 204148 54418 594732 202023 295515 368212 219782 239356 6000 96636 -49743 336518 208341 60469 597273 205848 299361 370437 224210 243295
Appendix IV Tables of Thermodynamic Properties 242
Table IV-2 Specific heat of some combustion products Enthalpy Reference Temp=298.15 Specific heat at constant pressure (J/mol/K) Standard state Pres. = 0.1Mpa
Source JANAF Thermochemical Tables (3rd Edition) T (K) CO CO2 H H2 H2O N N2 NO O O2 OH M 28.0104 44.0098 1.00794 2.01588 18.01528 14.0067 28.0134 30.0061 15.9994 31.9988 17.00734
100 29.104 29.208 20.786 28.154 33.299 20.786 29.104 32.302 23.703 29.106 32.627 200 29.108 32.359 20.786 27.447 33.349 20.786 29.107 30.420 22.734 29.126 30.777 300 29.142 37.221 20.786 28.849 33.596 20.786 29.125 29.841 21.901 29.385 29.977 400 29.342 41.325 20.786 29.181 34.262 20.786 29.249 29.944 21.482 30.106 29.650 500 29.794 44.627 20.786 29.260 35.226 20.786 29.580 30.486 21.257 31.091 29.521 600 30.443 47.321 20.786 29.327 36.325 20.786 30.110 31.238 21.124 32.090 29.527 700 31.171 49.564 20.786 29.441 37.495 20.786 30.754 32.028 21.040 32.981 29.663 800 31.899 51.434 20.786 29.624 38.721 20.786 31.433 32.767 20.984 33.733 29.917 900 32.577 52.999 20.786 29.881 39.987 20.786 32.090 33.422 20.944 34.355 30.264
1000 33.183 54.308 20.786 30.205 41.268 20.786 32.697 33.987 20.915 34.870 30.676 1100 33.710 55.409 20.786 30.581 42.536 20.786 33.241 34.468 20.893 35.300 31.124 1200 34.175 56.342 20.786 30.992 43.768 20.786 33.723 34.877 20.877 35.667 31.586 1300 34.572 57.137 20.786 31.423 44.945 20.786 34.147 35.226 20.864 35.988 32.046 1400 34.920 57.802 20.786 31.861 46.054 20.786 34.518 35.524 20.853 36.277 32.492 1500 35.217 58.379 20.786 32.298 47.090 20.786 34.843 35.780 20.845 36.544 32.917 1600 35.480 58.886 20.786 32.725 48.050 20.786 35.128 36.002 20.838 36.796 33.319 1700 35.710 59.317 20.786 33.139 48.935 20.786 35.378 36.195 20.833 37.040 33.694 1800 35.911 59.701 20.786 33.537 49.749 20.787 35.600 36.364 20.830 37.277 34.044 1900 36.091 60.049 20.786 33.917 50.496 20.788 35.796 36.514 20.827 37.510 34.369 2000 36.250 60.350 20.786 34.280 51.180 20.790 35.971 36.647 20.826 37.741 34.670 2100 36.392 60.622 20.786 34.624 51.823 20.793 36.126 36.767 20.827 37.969 34.950 2200 36.518 60.865 20.786 34.952 52.408 20.797 36.268 36.874 20.830 38.195 35.209 2300 36.635 61.086 20.786 35.263 52.947 20.804 36.395 36.971 20.835 38.419 35.449 2400 36.740 61.287 20.786 35.559 53.444 20.813 36.511 37.060 20.841 38.639 35.673 2500 36.836 61.471 20.786 35.842 53.904 20.826 36.616 37.141 20.851 38.856 35.881 2600 36.924 61.647 20.786 36.111 54.329 20.843 36.713 37.216 20.862 39.068 36.075 2700 37.003 61.802 20.786 36.370 54.723 20.864 36.801 37.285 20.877 39.276 36.256 2800 37.083 61.952 20.786 36.618 55.089 20.891 36.883 37.350 20.894 39.478 36.426 2900 37.150 62.095 20.786 36.856 55.430 20.924 36.959 37.410 20.914 39.674 36.586 3000 37.217 62.229 20.786 37.087 55.748 20.963 37.030 37.466 20.937 39.864 36.736 3100 37.279 62.347 20.786 37.311 56.044 21.010 37.096 37.519 20.963 40.048 36.878 3200 37.338 62.462 20.786 37.528 56.323 21.064 37.158 37.570 20.991 40.225 37.013 3300 37.392 62.573 20.786 37.740 56.583 21.126 37.216 37.617 21.022 40.395 37.140 3400 37.443 62.681 20.786 37.946 56.828 21.197 37.271 37.663 21.056 40.559 37.261 3500 37.493 62.785 20.786 38.149 57.058 21.277 37.323 37.706 21.092 40.716 37.376 3600 37.543 62.884 20.786 38.348 57.276 21.365 37.373 37.747 21.130 40.868 37.486 3700 37.589 62.980 20.786 38.544 57.480 21.463 37.420 37.787 21.170 41.013 37.592 3800 37.631 63.074 20.786 38.738 57.675 21.569 37.465 37.825 21.213 41.154 37.693 3900 37.673 63.166 20.786 38.928 57.859 21.685 37.508 37.862 21.257 41.289 37.791 4000 37.715 63.254 20.786 39.116 58.033 21.809 37.550 37.898 21.302 41.421 37.885 4100 37.756 63.341 20.786 39.301 58.199 21.941 37.590 37.933 21.349 41.549 37.976 4200 37.794 63.426 20.786 39.484 58.357 22.082 37.629 37.966 21.397 41.674 38.064 4300 37.832 63.509 20.786 39.665 58.507 22.231 37.666 37.999 21.445 41.798 38.150 4400 37.869 63.588 20.786 39.842 58.650 22.388 37.702 38.031 21.495 41.920 38.233 4500 37.903 63.667 20.786 40.017 58.787 22.551 37.738 38.062 21.545 42.042 38.315 4600 37.941 63.745 20.786 40.188 58.918 22.722 37.773 38.092 21.596 42.164 38.394 4700 37.974 63.823 20.786 40.355 59.044 22.899 37.808 38.122 21.647 42.287 38.472 4800 38.007 63.893 20.786 40.518 59.164 23.081 37.843 38.151 21.697 42.413 38.549 4900 38.041 63.968 20.786 40.676 59.275 23.269 37.878 38.180 21.748 42.542 38.625 5000 38.074 64.046 20.786 40.829 59.390 23.461 37.912 38.208 21.799 42.675 38.699 5100 38.104 64.128 20.786 40.976 59.509 23.658 37.947 38.235 21.849 42.813 38.773 5200 38.137 64.220 20.786 41.117 59.628 23.858 37.981 38.262 21.899 42.956 38.846 5300 38.171 64.312 20.786 41.252 59.746 24.061 38.013 38.289 21.949 43.105 38.919 5400 38.200 64.404 20.786 41.379 59.864 24.266 38.046 38.316 21.997 43.262 38.991 5500 38.233 64.496 20.786 41.498 59.982 24.474 38.080 38.342 22.045 43.426 39.062 5600 38.263 64.588 20.786 41.609 60.100 24.682 38.116 38.367 22.093 43.599 39.134 5700 38.296 64.680 20.786 41.712 60.218 24.892 38.154 38.393 22.139 43.781 39.206 5800 38.325 64.772 20.786 41.806 60.335 25.102 38.193 38.418 22.184 43.973 39.278 5900 38.355 64.865 20.786 41.890 60.453 25.312 38.234 38.443 22.229 44.175 39.350 6000 38.388 64.957 20.786 41.965 60.571 25.521 38.276 38.468 22.273 44.387 39.423
Appendix IV Tables of Thermodynamic Properties 243
Table IV-1 and Table IV-2 list the absolute enthalpy and specific heats of 11
combustion species. The source is the JANAF thermochemical tables [38].
The absolute enthalpy is enthalpy relative to stable elements at the reference state
(0.1MPa, 298.15K). It was calculated as.
where is sensible enthalpy, is the reference temperature 298.15K, and is the
enthalpy of formation at the reference state.
Table IV-3 Properties of some fuels (various sources [30, 62, 107]) nOctane Methane Ethane Propane Methanol Ethanol Gasoline Gasoline2 Diesel
Formula C8H18 C1H4 C2H6 C3H8 C1H4O1 C2H6O1 C8.26H15.5 C7.76H13.1 C10.8H18.7 M(g/mol) 114.2222 16.0416 30.0674 44.0932 32.0406 46.0664 114.825 106.401 148.556
(J/mol) -208826 -74933 -84801 -104062 -201342 -235055 -200000 -200000 -350000 BP (K) 399.15 112.15 184.15 231.15 338.15 351.15 387.15 387.15 493.15 (J/mol) 34381 8181 14703 19048 35213 38512 43634 40432 39900 A1 38.03599 6.946013 3.908762 10.49418 13.07256 9.627878 -8.3736 12.5604 -0.83736 A2 0.592813 0.101864 0.183411 0.246922 0.119191 0.225725 0.632207 0.519163 0.879228 A3 -0.00023 -4.17E-05 -7.40E-05 -9.70E-05 -5.13E-05 -1.13E-04 -2.51E-04 -1.95E-04 -4.06E-04 A4 3.43E-08 6.42E-09 1.02E-08 1.46E-08 7.69E-09 1.89E-08 3.35E-08 2.47E-08 6.28E-08 B1 40.7229 64.5193 89.7211 109.3511 81.0948 112.3099 81.5531 81.5531 165.9032 B2 0.2825 0 0 0 0 0 0.5657 0.5657 0.6933
Table IV-3 lists properties of various fuels. The specific heat , in gas phase was
derived from various references [30, 62, 107] as polynomial functions of
temperature. Some massaging of the data was required to convert disparate units to
J/mol and to guess the values likely at high temperatures since the functions given
were typically only valid between 250 and 1200K. Likewise the liquid phase
specific heat , enthalpy of evaporation , boiling point and Lower Heating
Value (LHV) were obtained where available from [30, 107] and otherwise estimated
based on similar fuels. The molar mass M was calculated based on atomic formula.
The enthalpy of formation was calculated based on the LHV. The values used
for nine typical fuels are shown in Table IV-3. Other fuels can be added to the
database in the future if necessary.
Specific heats in gas and liquid phase are represented respectively by polynomial
functions as:
For gas, these are valid up to about 3000K.
Appendix IV Tables of Thermodynamic Properties 244
Enthalpy of formation at standard state (298.15K, 0.1MPa) was calculated as
where LHV is in units of J/mol and is based on gas phase reactants and products.
J/mol
J/mol
The average composition of the fuel is
The absolute enthalpy of the fuel may be calculated by
Table IV-4 lists eight reactions and the formula for calculation the corresponding
equilibrium constant.
Table IV-4 Equilibrium equations
Reaction Equilibrium constant
1
2
3
4
5
6
7
8
The thermodynamic property is defined as
where is the molar specific entropy at the reference pressure (0.1Mpa) and is
the absolute enthalpy. Table IV-5 lists the equilibrium constant calculated for each
reaction using data from the JANAF tables [38].
Appendix IV Tables of Thermodynamic Properties 245
Table IV-5 Equilibrium constants for selected reactions Enthalpy Reference Temp=298.15 Equilibrium Constant ln(K) Standard state Pres. = 0.1Mpa
Source data JANAF Thermochemical Tables (3rd Edition) Reactions as listed in Table IV-4
T (K) 1 2 3 4 5 6 7 8 100 -284.545 -328.876 44.3309 -89.5166 -583.597 -1121.84 -511.035 -213.950 200 -139.976 -159.696 19.7200 -43.0614 -285.452 -554.538 -250.170 -105.599 300 -91.6072 -103.059 11.4520 -27.4608 -185.722 -365.173 -162.917 -69.4185 400 -67.3224 -74.6700 7.34761 -19.6410 -135.708 -270.364 -119.159 -51.3152 500 -52.6913 -57.6163 4.92503 -14.9476 -105.622 -213.403 -92.8293 -40.4494 600 -42.8986 -46.2436 3.34505 -11.8238 -85.5184 -175.381 -75.2257 -33.2041 700 -35.8776 -38.1227 2.24509 -9.59954 -71.1288 -148.189 -62.6170 -28.0280 800 -30.5937 -32.0358 1.44206 -7.93788 -60.3171 -127.772 -53.1340 -24.1458 900 -26.4709 -27.3060 0.83505 -6.65105 -51.8939 -111.875 -45.7384 -21.1251
1000 -23.1632 -23.5266 0.36334 -5.62670 -45.1455 -99.1455 -39.8065 -18.7083 1100 -20.4499 -20.4383 -0.01167 -4.79256 -39.6163 -88.7199 -34.9403 -16.7306 1200 -18.1837 -17.8684 -0.31529 -4.10096 -35.0029 -80.0245 -30.8755 -15.0818 1300 -16.2621 -15.6971 -0.56498 -3.51831 -31.0943 -72.6607 -27.4276 -13.6869 1400 -14.6120 -13.8390 -0.77297 -3.02131 -27.7405 -66.3441 -24.4656 -12.4908 1500 -13.1797 -12.2315 -0.94827 -2.59211 -24.8305 -60.8656 -21.8930 -11.4538 1600 -11.9248 -10.8273 -1.09748 -2.21796 -22.2819 -56.0683 -19.6374 -10.5467 1700 -10.8162 -9.59061 -1.22567 -1.88926 -20.0308 -51.8328 -17.6431 -9.74599 1800 -9.82965 -8.49328 -1.33637 -1.59831 -18.0280 -48.0655 -15.8671 -9.03443 1900 -8.94605 -7.51314 -1.43291 -1.33874 -16.2346 -44.6925 -14.2753 -8.39780 2000 -8.15016 -6.63286 -1.51729 -1.10626 -14.6193 -41.6547 -12.8403 -7.82491 2100 -7.42950 -5.83782 -1.59169 -0.89633 -13.1567 -38.9050 -11.5395 -7.30671 2200 -6.77376 -5.11646 -1.65730 -0.70647 -11.8259 -36.4040 -10.3555 -6.83572 2300 -6.17486 -4.45910 -1.71577 -0.53371 -10.6102 -34.1188 -9.27267 -6.40588 2400 -5.62541 -3.85764 -1.76777 -0.37604 -9.49492 -32.0231 -8.27902 -6.01206 2500 -5.11963 -3.30527 -1.81435 -0.23133 -8.46827 -30.0942 -7.36347 -5.64991 2600 -4.65240 -2.79635 -1.85605 -0.09854 -7.52028 -28.3123 -6.51727 -5.31577 2700 -4.21977 -2.32599 -1.89378 0.02421 -6.64179 -26.6617 -5.73290 -5.00683 2800 -3.81764 -1.89029 -1.92735 0.13752 -5.82584 -25.1284 -5.00380 -4.71991 2900 -3.44314 -1.48515 -1.95798 0.24247 -5.06567 -23.7002 -4.32389 -4.45346 3000 -3.09354 -1.10789 -1.98565 0.34013 -4.35606 -22.3663 -3.68903 -4.20476 3100 -2.76623 -0.75556 -2.01068 0.43094 -3.69183 -21.1177 -3.09427 -3.97213 3200 -2.45918 -0.42577 -2.03341 0.51542 -3.06890 -19.9469 -2.53630 -3.75453 3300 -2.17070 -0.11671 -2.05399 0.59458 -2.48344 -18.8462 -2.01158 -3.55035 3400 -1.89888 0.17365 -2.07253 0.66877 -1.93231 -17.8099 -1.51753 -3.35855 3500 -1.64255 0.44715 -2.08970 0.73821 -1.41260 -16.8321 -1.05114 -3.17771 3600 -1.40046 0.70475 -2.10521 0.80333 -0.92169 -15.9086 -0.61045 -3.00758 3700 -1.17109 0.94806 -2.11914 0.86458 -0.45700 -15.0340 -0.19326 -2.84655 3800 -0.95374 1.17828 -2.13201 0.92234 -0.01681 -14.2051 0.20222 -2.69446 3900 -0.74739 1.39614 -2.14353 0.97654 0.40108 -13.4181 0.57768 -2.55026 4000 -0.55122 1.60273 -2.15396 1.02782 0.79792 -12.6705 0.93446 -2.41338 4100 -0.36458 1.79891 -2.16349 1.07642 1.17577 -11.9585 1.27434 -2.28343 4200 -0.18659 1.98533 -2.17192 1.12215 1.53556 -11.2800 1.59776 -2.16026 4300 -0.01683 2.16295 -2.17978 1.16531 1.87869 -10.6327 1.90671 -2.04271 4400 0.14535 2.33202 -2.18667 1.20636 2.20646 -10.0145 2.20133 -1.93080 4500 0.30061 2.49327 -2.19266 1.24543 2.51949 -9.42327 2.48325 -1.82410 4600 0.44908 2.64745 -2.19836 1.28226 2.81899 -8.85730 2.75276 -1.72216 4700 0.59144 2.79459 -2.20315 1.31716 3.10588 -8.31487 3.01102 -1.62500 4800 0.72782 2.93547 -2.20765 1.35054 3.38088 -7.79469 3.25867 -1.53184 4900 0.85891 3.07036 -2.21145 1.38219 3.64467 -7.29537 3.49632 -1.44281 5000 0.98507 3.19952 -2.21445 1.41209 3.89800 -6.81559 3.72428 -1.35751 5100 1.10617 3.32346 -2.21729 1.44055 4.14127 -6.35429 3.94331 -1.27579 5200 1.22287 3.44252 -2.21965 1.46764 4.37552 -5.91004 4.15411 -1.19747 5300 1.33519 3.55674 -2.22155 1.49363 4.60088 -5.48235 4.35699 -1.12203 5400 1.44355 3.66659 -2.22304 1.51821 4.81794 -5.06998 4.55212 -1.04995 5500 1.54800 3.77219 -2.22419 1.54169 5.02708 -4.67220 4.74049 -0.98025 5600 1.64897 3.87406 -2.22509 1.56403 5.22848 -4.28813 4.92186 -0.91357 5700 1.74651 3.97209 -2.22558 1.58536 5.42330 -3.91735 5.09704 -0.84936 5800 1.84063 4.06644 -2.22581 1.60563 5.61116 -3.55875 5.26626 -0.78746 5900 1.93193 4.15758 -2.22565 1.62499 5.79286 -3.21198 5.42947 -0.72791 6000 2.02009 4.24547 -2.22538 1.64358 5.96843 -2.87631 5.58735 -0.67060
246
. . . .
247
Appendix V
Method for Calculating Chemical Equilibrium of
Combustion Products
This appendix should be read in conjunction with section 3.3. Some inspiration for
the following method is due to C. Olikara and G. L. Borman in the FORTRAN
source code TPEQIL available for download from the University of Wisconsin-
Madison at http://www.erc.wisc.edu/modeling/zerod.php
The moles of atomic carbon, hydrogen, oxygen and nitrogen in the mixture are
determined by summing up the contribution from each species present in the
mixture. The fuel has composition
V-1 Equation set
Five mass balance equations are
where is the total species moles in the mixture.
Eight possible equilibrium equations are listed in Table 3-2 and are repeated here
(note, is in bar)
Appendix V Method for Calculating Chemical Equilibrium of Combustion Products 248
Using six of these equations, six unknown species concentrations can be eliminated
by writing them in terms of and
Further, the carbon mass balance can be combined with the equilibrium equation
to form
and
Then the problem reduces to the solution of four equations in four unknowns
The four equations are:
Appendix V Method for Calculating Chemical Equilibrium of Combustion Products 249
Total of all fractions
Hydrogen mass balance
Nitrogen mass balance
Oxygen mass balance
To solve this set of equations using the Newton-Raphson method, the partial
derivatives of each function must be found. Referring to the six equilibrium
equations above:
1)
4)
5)
6)
7)
8)
And from the expression for above
The solution of equations f1, f2, f3 and f4 is a challenging numerical problem because
the fractions of the three unknown chemical species and can differ by
many orders of magnitude, as can the equilibrium constants K1-8. Where early
iterations of the Newton-Raphson method resulted in a negative value for one of the
unknown fractions, this was detected and the variable in question was res-set to a
value 100 times less than the previously guessed value. This ensured stability.
Iteration was terminated when all of the primary variables
changed less than 0.1%.
Appendix V Method for Calculating Chemical Equilibrium of Combustion Products 250
V-2 Low Temperature solution
The lower temperature limit for the full equilibrium calculation was set to 1400K.
Below this temperature, minor species are present in such low fractions (less than 10-
10) as to be negligible. Thus low temperature combustion was assumed with lean side
species being CO2, H2O, N2 and O2 and rich side species being CO, CO2, H2, H2O and
N2. For both situations, moles of nitrogen (diatomic) was found as
For lean mixture the solution in moles is
If the value calculated for is negative, this indicates there is no excess oxygen, so
a rich mixture is in view. The solution is more involved. Setting to zero there
are four unknowns, so the equilibrium equation
is used to close the problem which can be reduced to a single quadratic in
Where
Then
Appendix V Method for Calculating Chemical Equilibrium of Combustion Products 251
V-3 Equilibrium Initial Approximation
The starting point for an initial guess is the solution for low temperature combustion
converted to mole fractions. Next, for the case of a rich mixture the fraction of
oxygen was estimated by using either
Or
And limiting the result to a maximum value of
Mixtures which were exactly stoichiometric were problematic, so one more step was
taken to remove this problem. The concentrations of the following four species were
updated
This level of initial guess has proved sufficient for evaluating mixtures ranging from
1400K-6000K and equivalence ratio of 0-2.5. The number of iterations in the main
Newton-Raphson routine ranged from 2-10 with an average of about 5.
Appendix V Method for Calculating Chemical Equilibrium of Combustion Products 252
V-4 Partial derivatives of mixture properties
In some thermodynamic calculations, it is useful to know the rates of change of
mixture properties with respect to one variable. For example, given that for a
mixture of species with mole fractions
Then
Similar expressions can be written for enthalpy .
To evaluate the partials of the gas constant write
Then
Other useful applications for the mixture partial derivatives can be found. For
instance, for a fuel with composition we can write the enthalpy of
combustion at a certain atmospheric condition as
Appendix V Method for Calculating Chemical Equilibrium of Combustion Products 253
Figure V-1 shows the heat release per kg of n-octane when burned at various
temperatures and equivalence ratios. Numerous interesting features are apparent.
For instance it is obvious that while ever the mixture is lean, more energy can be
released by adding more fuel. However there is a sharp drop in heat release as the
mixture becomes rich of stoichiometric, and the result of adding fuel is an
endothermic reaction. It is also interesting to note that at very high temperatures,
addition of fuel does not release much heat, if any. Note however that if the
combustion products cool, the extra energy will be realised as dissociated species re-
combine.
Figure V-1 Heat release of n-octane at various equivalence ratios and
temperatures
V-4-1 Calculating the partials of each species fraction
To evaluate these mixture property partial derivatives, we need the partials of each
species mole fraction , as well as the total moles . These can be calculated as
follows.
Each of the four mass balance equations can be differentiated with respect to the
variable of interest (either of )
-2.E+7
-1.E+7
0.E+0
1.E+7
2.E+7
3.E+7
4.E+7
5.E+7
0 0.5 1 1.5 2 2.5
1800K
2100K
2400K
2700K 3000K
3300K
3600K
3900K
Equivalence ratio
Mixture made from 75 parts N2, 15 parts O2 by mass and various parts n-octane (C8H18)
3
-dH/
dmfu
el (J
/kg)
1500K
1 bar
Appendix V Method for Calculating Chemical Equilibrium of Combustion Products 254
Thus
This set of four linear equations can then be solved simultaneously for the four
unknowns .
The value of LHS can be calculated from the dependant species
Note that since the equilibrium constants are functions of temperature (only),
their derivative WRT temperature can be found.
The partial derivatives
are already known from the preceding equilibrium calculation which used the
Newton-Raphson method.
Once the partials of the primary variables
are evaluated, the partials of the remaining species can be evaluated by back
substitution.
255
Appendix VI
Derivation of fundamental one dimensional unsteady
gas equation
This derivation has been adapted from that shown by Blair [24] and Earnshaw [44].
Consider a small element of gas that is impelled by a pressure wave from one
direction as shown in Figure VI-1 (see also Figure 4-2).
Figure VI-1 A fluid element influenced by a pressure wave
Initially, fluid particles are at position , and after some time , they are at position y.
Initially the fluid element has a length . The cross section of the flow remains
constant so the cross section A of the element remains constant. Initially, the
pressure in the infinitesimal element is , and after time the pressure is . The
pressure in element varies according to the volume, and the change in pressure is
assumed isentropic. Thus
(VI-1)
Differentiating this expression WRT gives
(VI-2)
The next relationship to be employed in the solution is the momentum equation:
t0
t
A
x1
Particle path lines
Small fluid element
x2
y2
y1 tim
e
position (x, y)
References 256
For a constant mass this reduces to
which for the fluid element can be written:
(VI-3)
where is the acceleration of a fluid particle or infinitesimal element.
Then (VI-2) can be combined with (VI-3)
(VI-4)
since the speed of sound for a perfect gas is
The integration of equation (VI-4) WRT time will yield the velocity
To integrate this equation, first suppose the solution is of the form:
(VI-5)
Differentiating this equation can WRT gives (after some manipulation)
(VI-6)
This result can be substituted back into equation (VI-4) to give
References 257
Now integrating WRT
where is the constant of integration.
Making two further substitutions using equations (VI-1) and (VI-5) gives
There are two possible results here because we have not had to assume a direction
for the pressure wave.
At the pressure is and the velocity is . Solving the for gives
Thus, the equation for particle velocity is:
(VI-7)
The positive case corresponds to a wave travelling in the positive direction. The
negative case corresponds to a wave moving in the negative direction.
258
. . . .
259
Appendix VII
Derivation of Boundary Flow Equations
The flow at cell boundaries can be calculated as described in section 4.4. Figure
VII-1 shows a typical cell boundary (which could also be a duct boundary). In this
case there are five unknowns – pressure ( ), velocity ( ) and downstream
reference sound speed ( ). However other boundary flows will have various other
combinations of known and unknown quantities. The terms in the equations derived
below will follow the nomenclature of Figure VII-1 where the flow is from 1 2.
Figure VII-1 Schematic of a duct boundary
VII-1 Energy equation
Conservation of energy for a steady flow from 1 2 with gravitational potential
neglected is
Assuming calorically perfect flow this becomes
Position
time
Reflectedwaves Xr
Incident wavesXi
a Ra a0a b Rb
a0b
c Rc a0c
d Rd a0d
a0a a
a0b b
X1 X2 a01 a02
u1 u2
A1 A2
Xia Xib
Unknown values in bold. Flow is from left to right
Appendix VII Derivation of Boundary Flow Equations 260
From the definition of isentropic reference temperature
and writing
and since for a perfect gas
the energy equation becomes
VII-2 Continuity equation
Mass continuity for steady flow from 1 2 is
From the ideal gas relation
Then the mass flow rate is
The continuity equation becomes
Appendix VII Derivation of Boundary Flow Equations 261
VII-3 Wave equation
Gas property discontinuities exist (in general) immediately upstream and
downstream of the boundary flow as sketched Figure VII-1. Since it is a contact
surface between two regions of gas, there is no mass flow across this surface. It
follows therefore, that the contact surface moves with the local fluid and the velocity
of the fluid on either side immediately adjacent to the contact surface is equal.
Further, since the acceleration of the local fluid is finite, as the thickness of the
contact surface approaches zero, the pressure difference across the surface
approaches zero. Thus
And
Writing in terms of the pressure amplitude ratio
Combining the pressure wave equations (4-3) and (4-4) and eliminating one of the
pressure waves we can write for the cell space on each side of the boundary flow
Substituting and for and above gives
Appendix VII Derivation of Boundary Flow Equations 262
VII-4 equation
For thermal energy added to the fluid, the change in temperature (for constant
pressure) is proportional to the energy as
Where is any form of frictional dissipation. Writing
the change in temperature due to friction work is
Writing in terms of isentropic reference temperature
Where is at the pressure at which the thermal energy is being added to the fluid.
263
Appendix VIII
Model Data Structures
This section lists the data structures that were used to form the model. There are
three main categories of object that make an engine model:
Ducts – These are the gas dynamic parts of the model. They are given a length and
cross section, which may be tapered. A duct is made of n cells with n+1 cell
boundaries or nodes. Each end of the duct may be connected to another duct or
volume. If the end is unconnected, the boundary is closed.
Volumes – These are the parts of the model where gas dynamic effects don’t need to
be modelled, such as the atmosphere and cylinder. Volumes are typically larger in
cross section than ducts, and are modelled as zero dimensional thermodynamic
control volumes. A large reservoir such as the atmosphere can be modelled as an
infinitely large volume which supplies a steady pressure to any connecting ducts.
Bodies – the moving parts of the model are represented by bodies. These may be
parts such as pistons and valves. The Pempek free-piston engine had several floating
bodies which interacted through spring forces, contact and friction, such as the
mover and piston mounted passive inlet valves. Bodies can be specified in the
model to constrain the length of certain ducts, determine the volume of certain
volumes (such as the cylinders), or specify the geometric flow area through valves.
Some other data structures control how the above model objects are connected to one
another:
Connect structure – This manages the flow connections between ducts and volumes.
It also stores some flow data at these connections for convenient retrieval.
Body Interaction structure – This manages how the various bodies interact with one
another (if at all). It contains information on the kinds and locations of collisions
possible between each body, the relative distance related forces such as linear spring
force, and relative velocity forces such as kinetic friction.
The code was programmed in Matlab, which implements a convenient data type
“Structure” which uses labelled fields which may contain arrays of differing data
class and size. This is especially useful in the duct structure since the number of
cells may change halfway through a simulation due to automatic re-meshing. Thus
the data structures shown here may need to be modified if coded in a different
language.
Appendix VIII Model Data Structures 264
Duct data structure
D(d) contains details and flow history for each duct d in the model .name ‘string’ convenient label .description ‘string’ .enabled(t) [Y/N Y/N . . . ducts may be temporarily disabled .friction_function I specifies if and how friction is calculated .heat_transfer_function I specifies if and how heat transfer is calculated .combustion function I specifies if and how combustion is calculated .re_mesh_function I specifies if and how the duct is re-meshed .flow_seperation I specifies if and how flow separation is calculated .A_linear Y/N allows annular ducts to be modelled .mixing 0-1 specifies degree of mixing .conserve_mass Y/N specifies whether mass conservation is enforced .user ? field for customised storage of date .position(1:2) .body I duct end location .offset #.# duct end offset from above location .contributes_to_wall_velocity Y/N
.c(t) the cell based data for the duct at each timestep t .A(c) [# # . . . cross sectional area of each cell .C(c) [# # . . . perimeter of surface of each cell .Cp(c) [# # . . . Specific heat of gas in each cell .Cv(c) [# # . . . Specific heat of gas in each cell
.m(s,c) mass of each species in each cell
.T0(c) [# # . . . reference temperature in each cell Tw(c) [# # . . . wall temperature of each cell
.v_wave(1:2,c) rightward and leftward wave velocity in each cell
.ave_fastest_wave_traverse # duct average of the fastest wave in each cell .dq_dt(c) [# # . . . specific heat transfer rate at each cell .dT0(c) [# # . . . change in ref. temp. due to heat transfer .P_corr(c) [# # . . . pressure correction applied in each cell .m_corr(c) [# # . . . mass correction applied in each cell .dX(c) [# # . . . pressure wave change from heat trans. and P corr .A_end(1:2) [# #] cross sectional area of each end of the duct .C_end(1:2) [# #] perimeter of the surface of each end of the duct
.ket(1:2,c) turbulent KE (for heat transfer model)
.n(t) the node based data for the duct at each timestep t .P(1:2,n) pressure either side of each node
.c(1:2,n) flow velocity either side of each node
.Xi(1:2,n) incident pressure wave on each side of each node
.Xr(1:2,n) reflected pressure wave on each side of each node
.x(n) [# # . . . position of each node .v(n) [# # . . . velocity of each node (if duct is moving)
.dm_dt(s,n) mass flow rate of each species at each node
.T0(1:2,n) reference temp. at each side of each node
.ket(1:2,n) turbulent KE at each side of each node (heat transfer model)
Appendix VIII Model Data Structures 265
Volume data structure
V(v) contains details and flow history for each volume v in the model .name ‘string’ convenient label .description ‘string’ .enabled(t) [Y/N Y/N . . . volumes may be temporarily disabled .CSA # effective cross s. area. For volume .length details for determining the effective lengh of the vol. .body [# #] up to two bodies to specify effective length of vol. .offset # constant offset to above length .heat_transfer_function I specifies if and how heat transfer is calculated .combustion_function I specifies if and how combustion is calculated .user ? field for customised storage of date .P(t) [# # . . . pressure at each time step .T(t) [# # . . . temperature at each time step .dT_dt(t) [# # . . . temp rate of change at each time step .Cp(t) [# # . . . specific heat at each time step
.Cv(t) [# # . . . specific heat at each time step .Tw(t) [# # . . . wall temp rate of change at each time step .H(t) [# # . . . total enthalpy flow rate at each time step .dq_dt(t) [# # . . . specific heat transfer rate at each time step
.m(s,t) mass of each main species at each time step
.extra_m(s,t) mass of any minor species at each time step
.dm_dt(s,t) mass flow of each main species at each time step.
.dm_ch(s,t) chemical mass rt. of ch. of each species at each time st.
.u(s,t) specific internal energy for each species at each time st.
Appendix VIII Model Data Structures 266
Body data structure
B(b) contains details and trajectory history for each body b in the model .name ‘string’ convenient label .description ‘string’ .enabled(t) [Y/N Y/N . . . volumes may be temporarily disabled .user ? field for customised storage of date .m # mass of body .x(t) [# # . . . position of body at each timestep .v(t) [# # . . . velocity of body at each timestep .a(t) [# # . . . acceleration of body at each timestep .body_position specifies the method for calculating the body trajectory .index I .Fp specifies any force on the body not due to another body .index I specifies the method for calculating force User_inputs(n) [# # . . . any necessary details .f(t) [# # . . . non-body-body force history
Body Interaction structure
I(b1,b2) contains details and force history for each interacting pair of bodies .upper_collision Y/N .upper_collision_offset # .upper_collision_k 0-1 .lower_collision Y/N .lower_collision_offset # .lower_collision_k 0-1 .Fx(n) specifies up to n different spring type forces between the bodies .index I specifies the force calculation function .user_inputs [# # . . . various necessary details .f # force at current timestep .Fv(n) specifies up to n different friction type forces between the bodies .index I specifies the force calculation function .user_inputs [# # . . . various necessary details .f # force at current timestep
Appendix VIII Model Data Structures 267
Flow connections structure
C(a,b) contains details and flow history for each flow connection in model .connection Y/N flags the presence of a connection .junction_mode_Cl # specifies the loss coefficient for multi-pipe junctions .volume_velocity # specifies any velocity that would raise the dynamic P .A details for calculating the flow CSA of the connection .body [# #] up to two bodies to specify the flow area .body_offset # offset for the above length .multiplier # multiplier for the opening .Ca I flow area coefficient map to look up .entry Y/N specifies forward or reverse flow .reference # reference area .PR # pressure ratio .flow(t) records history details of the connection flow .c # velocity .P # pressure .T0 # reference temperature .T # temperature .Cp # specific heat .Cv # specific heat
.dm_dt(s) mass flow rate of each species
Flow area coefficient structure
Ca(ca) contains data for specifying the flow area coefficient for certain boundary flows .name ‘string’ convenient label .description ‘string’ .data(1:2) contains the data for forward and reverse flow area coefficients .raw data points from steady flow experiments
.AR [# # . . . area ratio points for fitted data .PR [# # . . . pressure ratio points for fitted data .Ca_fit(PR,AR) net of fited falues corresponding to .AR and .PR
.surf Y/N specifies a surface or line .interp_method ‘string’ method for interpolating Ca_fit ‘linear’ or ‘cubic’
268
. . . .
269
Appendix IX
2nd Order Interpolation – further details
The gas dynamics model described in Chapter 4 is made of a string of idealised
segments (cells) which are constant area, constant property and frictionless. At the
connection point of each of these idealised segments (nodes), any area change, gas
property change or friction is accounted for. Therefore, the pressure and velocity of
the modelled flow usually has some step change across each node, as the effects of
area change, property change and friction modify the flow at this point. The general
situation is illustrated in Figure IX-1, which shows a typical variation in the
rightward pressure wave along a duct at a certain instant in time. The rightward
pressure wave is made up of incident and reflected waves (as described in Figure 4-
9) and the wave’s value is discontinuous at node 2.
Figure IX-1 Interpolating discontinuous pressure waves
The second order interpolation method requires the wave’s value to be known at
three points along the curve. In the example above, the propagation of the rightward
wave is being evaluated – that is, the future value of the wave incident on position
(node) 3 is being evaluated. The values of the three data points are:
The value of the wave at the far ‘upwind’ node (1) is modified for the purpose of the
interpolation calculation to remove the effect of the flow discontinuity at node 2. A
similar procedure is used for the leftward wave.
Rightward wave value
XR
Xr1
position, node
Xr2 Xi2 Xi3
1 2 3
Appendix VIII Model Data Structures 270
A special problem occurs for the node located second from the end of each duct. In
this case, no third ‘upwind’ node exists (as it lies outside the duct boundary). The
situation is illustrated in Figure IX-2, where the wave incident on node 2 at the
current time step has no third ‘upwind’ node at the previous time step with which to
perform the interpolation.
Figure IX-2 Handling the duct ends
A fake third upwind node (0) is created by using the current time step wave value at
node 1 (which is the duct’s left hand end). This method depends on the duct ends
being evaluated prior to this step, so that legitimate wave values at the duct ends are
already known. The position of the fake upwind node is set to the distance the wave
at this location travels in one time step.
If the duct is very short and only has one cell, then this technique cannot work. This
is because the duct ends are also the problematic ‘second from the end’ nodes. This
is illustrated in Figure IX-3. One of two options may be taken here. The
interpolation may revert to a first order (linear) method with the risk of numerically
unstable extrapolation if the courant number is greater than one, or the fake third
upwind node could be set to the value of the end node (at previous time step), which
should be stable, even at courant numbers greater than one.
Figure IX-3 Single cell ducts
time
previous time-step
Right travelling wave
position, node 1 2 3 4
current time-step
0
time
previous time-step
position, node 1 2
current time-step
271
Appendix X
Re-meshing Criteria and Method
Re-meshing criteria
Automatic mid-simulation duct re-meshing allows individual ducts to maintain
Courant numbers closer to unity. This aim is assisted by the second order wave
interpolation method (see section 4.3.1 and Appendix IX) which allows Courant
numbers somewhat greater than unity.
The process of re-meshing a duct introduces small interpolation errors in all flow and
fluid variables, so the frequency of re-meshing should be minimised. An effective
re-meshing criterion was developed and is described below.
Each cell has a right and left travelling wave which, if there is any local flow
velocity, are propagating at differing speeds. Thus each cell has two Courant
numbers, one for the leftward wave and one for the rightward wave. The higher of
these two Courant numbers are averaged with all the cells in the duct to give the
average Courant number. Also, the highest Courant number for the whole duct is
determined.
The duct is analysed for re-meshing if:
the average Courant number is greater than 1.1 or
the highest Courant number is greater than 1.2 or
the average Courant number is less than 0.85
The optimum number of cells for the duct is then calculated using three methods:
1. the fewest cells which would give an average Courant number greater than 1
2. the most cells which would give an average Courant number less than 1.05
3. the most cell which would give a highest courant number less than 1.15
The result which gives the fewest number of cells (ie the lowest Courant number) is
chosen. Note that the result may be exactly the same as the existing number of cells
in which case there is no change (no-re-meshing) of the duct.
Appendix X Re-meshing Criteria and Method 272
Re-meshing method
In the mesh implementation used here, pressure, velocity, temperature and species
fraction are stored at nodes, while all other duct properties are stored as cell values.
Figure 4-7 is repeated below, showing the nodes as solid circles, and the cell centres
as hollow circles. Note that the cell centres are defined midway between the nodes
both in time and space.
Figure 4-7 Re-meshing a duct
Re-meshing occurs after the most recent cell properties have been evaluated but
before the wave propagation from previous time step to current time step is
calculated. Thus re-meshing involves interpolating nodal values at the previous time
step and cell values that are most recent. Nodes at the current time step are also set
up; however the flow data here are yet to be determined in the calculation sequence
of the model (see section Chapter 6), so no interpolation is necessary.
The nodal and cell data are interpolated using a suitable method, in this case a high
order monotonic fitted curve provided by the MatLab function ‘pchip’. Note that
nodal pressure, velocity and temperature are all defined on both sides of each node,
so there are a total of 6 variables here that need to be interpolated. See Appendix
VIII for a full list of variables for nodes and cells.
time
current time-step
previous time-step
position
273
Appendix XI
Single Shot Experiments Cross Reference
Table XIV-1 lists the corresponding figure numbers in the original publication by
Kirkpatrick [68] for the single shot tests in sections 7.2 to 7.6
Table XIV-1 Cross reference figure numbers for single shot data
Figure number Corresponding figure in [68] Figure number Corresponding
figure in [68] Figure 7-10 Figure 6-14 Figure 7-41 Figure 6-108 Figure 7-11 Figure 6-187 Figure 7-42 Figure 6-168 Figure 7-12 Figure 6-191 Figure 7-43 Figure 6-109 Figure 7-13 Figure 6-18 Figure 7-44 Figure 6-169 Figure 7-14 Figure 6-189 Figure 7-45 Figure 6-56 Figure 7-15 Figure 6-193 Figure 7-46 Figure 6-126 Figure 7-16 Figure 6-26 Figure 7-47 Figure 6-57/127 Figure 7-17 Figure 6-32 Figure 7-48 Figure 6-58 Figure 7-18 Figure 6-7 Figure 7-49 Figure 6-128 Figure 7-19 Figure 6-10 Figure 7-50 Figure 6-59/129 Figure 7-20 Figure 6-162 Figure 7-51 Figure 6-64 Figure 7-21 Figure 6-164 Figure 7-52 Figure 6-134 Figure 7-22 Figure 6-168 Figure 7-53 Figure 6-65/135 Figure 7-23 Figure 6-8 Figure 7-54 Figure 6-66 Figure 7-24 Figure 6-11 Figure 7-55 Figure 6-136 Figure 7-25 Figure 6-15 Figure 7-56 Figure 6-67/138 Figure 7-26 Figure 6-18 Figure 7-57 Figure 6-137 Figure 7-27 Figure 6-175 Figure 7-58 Figure 6-139 Figure 7-28 Figure 6-178 Figure 7-59 Figure 6-140 Figure 7-29 Figure 6-179 Figure 7-60 Figure 6-141 Figure 7-30 Figure 6-180 Figure 7-61 Figure 23 1 Figure 7-31 Figure 6-181 Figure 7-62 Figure 20 1 Figure 7-32 Figure 6-182 Figure 7-63 Figure 6-87 Figure 7-33 Figure 6-183 Figure 7-64 Figure 6-154 Figure 7-34 Figure 6-184 Figure 7-65 Figure 6-88/155 Figure 7-35 Figure 6-185 Figure 7-66 Figure 6-90 Figure 7-37 Figure 6-96 Figure 7-67 Figure 6-156 Figure 7-38 Figure 6-160 Figure 7-68 Figure 6-91/157 Figure 7-39 Figure 6-97 Figure 7-69 Figure 23 [27] Figure 7-40 Figure 6-161 Figure 7-70 Figure 20 [27]
1 This figure from reference [27] Blair, G. P., Kirkpatrick, S. J., Mackey, D. O., and Fleck, R., 1995, Experimental Validation of 1-D Modelling Codes for a Pipe System Containing Area Discontinuities, SAE, Paper 950276
274
. . . .
275
Appendix XII
Derivation of normal shock equations
The derivations presented here are adapted from the derivations presented by
Anderson [13].
Consider the shock shown in Figure XII-1. The shock is perpendicular (normal) to
the direction of the flow. A control volume is constructed around the shock so that
the flow through the control volume is steady.
At the shock there is a change of velocity u, pressure P, temperature T, density and
local speed of sound a. The flow velocity u is defined relative to the shock.
The flow across the shock obeys conservation of energy, mass and momentum. If
the flow is calorically perfect, then an algebraic solution is available.
Figure XII-1 Schematic of a normal shock
The equation for the conservation of energy for a steady, adiabatic flow with no
work, between two states is
If the flow is calorically perfect we can write
and writing
Stationary Shock
Control volume Flow direction
Appendix XII Derivation of normal shock equations 276
and since for a perfect gas
Then
(XII-1)
We now define a state where the flow velocity equals the local sonic velocity
. This is a hypothetical state representing a flow that has undergone some adiabatic
process to accelerate or decelerate it to sonic velocity. The energy equation for this
situation can be expressed as:
Writing the hypothetical sonic flow as a function of the general state and .
(XII-2)
Alternatively:
(XII-3)
Equations (XII-2) and (XII-3) are expressions of the energy equation for a steady
adiabatic flow. Note that no isentropic assumption has been made.
The next step in the solution is to introduce both the conservation of mass and the
conservation of momentum equations, which must hold true for the control volume
of Figure XII-1.
Continuity equation can be written here as:
(XII-4)
The momentum equation is
For the steady flow of Figure XII-1 this can be written
Appendix XII Derivation of normal shock equations 277
Then the momentum equation can be written as
(XII-5)
Or since
and since for a perfect gas
The continuity/momentum equation can be written as:
Substituting equation (XII-3) for and for gives:
This expression can be simplified to:
(XII-6)
With this simple relation we are now able to find the flow on one side of the shock in
terms of known values on the other side. The expressions developed below assume
the upstream state 1 is known, and the downstream state 2 is being calculated, but
the reverse can also be done (that is, calculating the conditions upstream of a shock
based on the downstream conditions).
One further restriction that is not built into the equations here, but is a condition of
second law of thermodynamics, is that the only physically valid shock is one where
the flow is being compressed by the shock. Thus in Figure XII-1 the velocity of the
upstream flow must be greater than the local sonic velocity. The effect of the
shock will then be to slow, compress and heat the flow so that it is sub-sonic on the
downstream side.
Combining equation (XII-6) with equation (XII-2) yields the downstream velocity
as:
Appendix XII Derivation of normal shock equations 278
(XII-7)
The downstream pressure can now be found by rearranging the momentum equation
(XII-5)
And recalling
so that
(XII-8)
Or substituting away with equation (XII-7)
(XII-9)
Finding the remaining downstream values is straight forward, for instance from the
ideal gas relation
Using the continuity equation (XII-4)
(XII-10)
Then
(XII-11)
And
Appendix XII Derivation of normal shock equations 279
(XII-12)
Alternatively, if the speed of the shock is unknown but the pressure ratio across the
shock is known, the shock speed can be calculated by re-arranging equation (XII-
9)
(XII-13)
The downstream velocity can now be found by employing equation (XII-7)
(XII-7)
or by writing equation (XII-8) as
(XII-14)
The change in velocity across the shock in terms of the pressure ratio can be found
by writing equation (XII-14) as
And substituting in equation (XII-13)
(XII-15)
The normal shock equations derived here can be used for the case of a moving shock
by examining Figure XII-2 and noting that the shock velocity relative to the pre-
shocked gas is
and
Appendix XII Derivation of normal shock equations 280
Then the velocity of the post shock gas v is
v v
v
Figure XII-2 Travelling shock
Pre shock Post shock
+ve direction Travelling Shock
281
Appendix XIII
Rayleigh and Fanno Flow
XIII-1 One dimensional flow with heat transfer (Rayleigh
flow)
The momentum equation for frictionless flow in a constant area duct is given by
equation (XII-5)
Since according to the continuity equation
and for a perfect gas
The momentum equation can be written
It is useful to use the Mach number, defined as
Thus
(XIII-1)
Or alternatively simply
or
From the ideal gas equation and the continuity equation
Appendix XIII Rayleigh and Fanno Flow 282
From the definition of Mach number and since
thus
(XIII-2)
And again
(XIII-3)
For a flow from the total heat per kilogram added to the flow can be found by
applying the energy equation (XII-1)
or
(XIII-4)
If heat transfer is specified in the problem, along with flow conditions at a certain
point, then the solution can be found by plotting the flow as a function of Mach
number or velocity using equations (XIII-1) to (XIII-4), and locating the Mach
number where heat transfer is equal to that specified. This is a trial and error
approach. Note that the maximum heat that can be added to a flow will be when the
resulting flow is sonic.
Appendix XIII Rayleigh and Fanno Flow 283
XIII-2 One dimensional flow with friction (Fanno flow)
The energy equation for steady adiabatic flow of a perfect gas equation (XII-1) is
The Mach number is defined as
Thus
(XIII-5)
Or alternatively simply
From the ideal gas equation and the continuity equation
From the definition of Mach number and since
Then
(XIII-6)
And again
Appendix XIII Rayleigh and Fanno Flow 284
(XIII-7)
For a flow from with friction the total shear force on the wall can be found by
applying the momentum equation
For constant area flow with friction this becomes
Where is the friction force and is directed opposite to the flow.
Thus
And noting
then
(XIII-8)
If friction force is specified in the problem, along with flow conditions at a certain
point, then the solution can be found by plotting the flow as a function of Mach
number or velocity using equations (XIII-5) to (XIII-8), and locating the Mach
number where friction is equal to that specified. Since friction force is generally a
function of flow velocity an iterative approach may be needed. Note that the
maximum friction force that can be imposed on a flow will be when the resulting
flow is sonic. Thus it is a physical impossibility for a flow that is subsonic to driven
supersonic by friction alone, and likewise for an initially supersonic flow to be
driven subsonic by friction alone.
285
Appendix XIV
Graphical user interface screen shots
To assist in creating engine models, a graphical user interface was created. This
allowed individual engine model creation to be carried out quickly and accurately. It
also facilitated rapid and accurate editing of models to test the results of design
changes. A further extension of the graphical user interface was for post processing
the simulation results. This was necessary given the large volume of time-series data
contained in a single simulation.
Figure XIV-1 shows the main screen which is used to load saved models, edit basic
simulation parameters (such as time step number and size), and launch the model
editing screens. Figure XIV-2 shows the Duct creation/editing screen, where the
parameters of each duct in the model are set. Figure XIV-3 and likewise shows the
Volume editing screen. Connections between ducts to ducts, volumes to volumes
and volumes to ducts are also specified here and can be set from either screen.
Figure XIV-4 shows the Body editing screen. This screen also allows the user to
specify the relationships between bodies (if any) such as collisions, friction, spring
forces and user customisable forces (such as generator force). Figure XIV-5 shows
the Flow Coefficients screen. This allows the user to add and edit flow area
coefficients for various situations. This database also contains valve aerodynamic
force coefficients that are used to calculate the dynamics of the inlet and exhaust
valves. Figure XIV-6 shows the Function editing screen. This allows the user to add
and edit the arguments and return variables for a range of customisable functions.
Figure XIV-7 shows the screen for setting up animation plots. Six example
animations are included in the data CD that is packaged with this thesis, and are
listed in Appendix XV. Figure XIV-8 shows and example animation screen grab.
The data series at any time step can be exported to the clip-board. Figure XIV-9
shows the time-series plotter. This screen allows any time series data from the
model to be plotted, and optionally exported to the clip-board.
Appendix XIII Rayleigh and Fanno Flow 286
Figure XIV-1 Main screen
Figure XIV-2 Edit Ducts screen
Appendix XIII Rayleigh and Fanno Flow 287
Figure XIV-3 Edit Volumes screen
Figure XIV-4 Edit Bodies screen
Appendix XIII Rayleigh and Fanno Flow 288
Figure XIV-5 Edit Area Coefficients screen
Figure XIV-6 Edit Functions screen
Appendix XIII Rayleigh and Fanno Flow 289
Figure XIV-7 Create Animated plot screen
Figure XIV-8 Example Animation screen grab (shock tube problem)
Appendix XIII Rayleigh and Fanno Flow 290
Figure XIV-9 History plot screen
291
Appendix XV
Table of contents of data CD
Thesis.pdf Thesis soft copy
points.mp4 3CFD scavenging animation
existing engine motored.mp4 1D Motored engine simulation (see section 7.7)
existing engine fired.mp4 1D Fired engine simulation (see section 7.7)
modified engine 22mg.mp4 1D Fired simulation of modified passive inlet valve
layout with lower compressor pressure and tuned
exhaust pipe (see section 8.1)
port scavenged 2_5mg.mp4 1D port scavenged simulation, 2.5mg fuel injection
(see section 8.3)
port scavenged 5mg.mp4 1D port scavenged simulation, 5mg fuel injection
(see section 8.3)
port scavenged 22_5mg.mp4 1D port scavenged simulation, 22.5mg fuel
injection (see section 8.3)